158 results for “LMS”

TextSeal2 is a distortion-free LLM text watermark that merges two or more keys into one virtual key and couples it to fresh noise through AutoMark's Gaussian copula. Under speculative decoding it verifies keyed drafts with an extra exponential race, so output matches the target model and acceptance rate is unchanged. On three LLMs at a 10^-3 false-positive rate, it detects 8 to 22 points more 100-token answers than TextSeal.

arxiv.orgpaperAdded 31 Oct 2026, published 29 Oct 2026🛡️ Security

TextSeal2 is a distortion-free watermark for LLM output that keeps its signal under speculative decoding. It merges two or more keys into one virtual key coupled to fresh noise through AutoMark's Gaussian copula, and verifies drafts with an extra exponential race. On three LLMs it detects 8 to 22 points more 100-token answers than TextSeal at a 10^-3 false-positive rate.

arxiv.orgpaperAdded 31 Oct 2026, published 29 Oct 2026🧠 AI

PhysCount is a benchmark of 546 real-world images and 4,595 human-written questions in 651 target-specific groups. It tests whether multimodal LLMs can separate visible appearances (direct view, reflections, glass) from the physical entities behind them. Across 40 MLLMs, counting appearances is much easier than linking them to entities, and the best model answers every question correctly in only 13.36% of groups.

arxiv.orgpaperAdded 31 Oct 2026, published 29 Oct 2026🧠 AI

Understanding and Mitigating Role Spoofing in LLMs is an arXiv paper (October 2026) on role spoofing, a prompt injection where malicious tool output impersonates the user. It shows models track "who said what" using both content cues (semantics, style) and structure cues (control tokens, formatting). Attacks combining fake structure and content succeed often, adversarial training stops only content-based attacks, and linear probes trained to predict conversation roles detect spoofing across settings.

arxiv.orgpaperAdded 31 Oct 2026, published 29 Oct 2026🛡️ Security

A five-task benchmark tests multimodal LLM color judgment against CIE colorimetric ground truth: AWB pair judgment, skin-tone bias direction, CCT estimation, Delta E 2000 pair ordering, and color naming under changing illuminants. Eight endpoints were tested. On ColorChecker-24 stimuli, closed frontier models reach CCT MAE of 715-1056 K versus 1358-2133 K for open-weight models, and a non-ColorChecker control shows this gap is partly template-dependent.

arxiv.orgpaperAdded 31 Oct 2026, published 29 Oct 2026🧠 AI

TrussTutor-Bench is a benchmark of 200 verified two-dimensional truss problems from statics, paired with automatically generated student solutions. It evaluates thirteen LLMs on solution accuracy (64.5% to 97.0%), error detection (90.4% to 100%) and sycophancy. Models rejected their own correct solutions in 1.6% to 50.0% of cases when a student insisted the work was wrong. Sycophancy separated models; error detection did not.

arxiv.orgpaperAdded 31 Oct 2026, published 28 Oct 2026🧠 AI

This paper studies whether vision-language models can infer how long a person stays somewhere (living, staying, or leaving) from geotagged social media posts. The authors build a new corpus of geotagged tweets co-annotated by humans and LLMs. State-of-the-art models struggle even with multi-turn Chain-of-Thought prompting, often missing contextual cues and subtle hashtags while over-relying on prominent visual details.

arxiv.orgpaperAdded 30 Oct 2026, published 28 Oct 2026🧠 AI

inkGround is a benchmark for stroke trajectory recovery from full-page, camera-captured handwriting, with 537 page images (about 18.7k words, 163k strokes) and pseudo-label references filtered by OCR agreement and human raters. Models get a page image and a target word, then output ordered stroke points. Seven VLMs were scored with DTW, Chamfer distance, readability and style similarity. Closed-source models, notably Gemini-3.1-Pro, did best, and the ranking held on IAM-OnDB.

arxiv.orgpaperAdded 30 Oct 2026, published 28 Oct 2026🧠 AI

LongChang is an automated framework that builds simulated worlds with altered physical laws, such as heavier objects falling more slowly, to test whether LLMs can find the law by free exploration. Explorer LLMs manipulate objects and build tools without tasks or hints, while separate LLM agents generate scenes, review tool designs and grade reports. Across 12 LLMs, GPT-6 Astra reached a 16.7% discovery rate and the other 11 scored 0%.

arxiv.orgpaperAdded 28 Oct 2026, published 27 Oct 2026🧠 AI

VideoLVR is a method that lets video multimodal LLMs build latent visual evidence from local regions at relevant moments and reason over it interleaved with text tokens. A visual token prediction objective, defined over a vocabulary derived from the input video, supervises the latents. It extends to reinforcement learning with sampled visual tokens as actions and final-answer rewards, and improves fine-grained video benchmarks without much context growth.

arxiv.orgpaperAdded 28 Oct 2026, published 27 Oct 2026🧠 AI

This paper defines a character naming task for films: given a name spoken in dialogue and its video context, the model must pick which visible character bears that name. It provides a dataset with annotations for character naming and first appearances, a video-and-dialogue model, and a comparison with proprietary multimodal LLMs. It also proposes a progressive framework that builds a character bank as the film plays, without cast lists or actor photographs.

arxiv.orgpaperAdded 28 Oct 2026, published 27 Oct 2026🧠 AI

This paper studies how multimodal LLMs refuse questions that an image cannot answer, using the MoHoBench benchmark on three MLLMs. It varies the meaning and placement of refusal instructions, compares supervised and preference-based post-training, and intervenes on a refusal-related activation direction. Instructions change refusal at the sample level in ways aggregate rates hide. Post-training raises refusal most but can mix refusal with attempted answers. Activation intervention gives smaller, transferable changes.

arxiv.orgpaperAdded 28 Oct 2026, published 27 Oct 2026🧠 AI

Falko Strenzke's ePrint 2026/2635 presents a multi-target second-preimage attack on message hashing in HSS/LMS and XMSS with SHA-256 and SHA-256/192. It herds chaining states from many collected signatures into one state, defeating the key and leaf-index prefix. With 2^64 signatures, SHA-256 cost falls from 2^256 to about 2^166 (2^156 at maximal message length). SHAKE256 parameter sets are unaffected.

eprint.iacr.orgpaperAdded 26 Oct 2026, published 25 Oct 2026🛡️ Security

Open bilingual (Russian and UK English) benchmark of five open pseudonymisation tools and a regex baseline, plus a new engine built on the two strongest tools. On a held-out set of 1,000 synthetic documents per language, the best existing tools leave a whole personal value visible in 61% (Russian) and 58% (UK) of documents. The engine does so in 0.3% and 0.2%. Russian name restoration reaches only 94.6%, below the 99% threshold. Real documents were not tested.

arxiv.orgpaperAdded 27 Oct 2026, published 25 Oct 2026🛡️ Security

This paper presents an open benchmark of five existing pseudonymisation tools and a regex baseline on Russian and UK English documents, plus an open engine built on the two strongest tools. On a held-out set of 1,000 synthetic documents per language, the best existing tools leave a whole value visible in 58 to 61% of documents. The engine does so in 0.2 to 0.3%. Russian name restoration reaches only 94.6%, below the 99% threshold. Real documents were not tested.

arxiv.orgpaperAdded 27 Oct 2026, published 25 Oct 2026🧠 AI

This paper presents a pipeline that uses persona-conditioned vision language models to surface interpretive ambiguity in visualizations. Topic-relevant personas elicit candidate takeaways, and a Target-Pattern-Scope framework lets LLM judges label statement pairs as Compatible, Disjoint or Opposing, which are summarized in an ambiguity profile. The authors benchmark VLMs on chart comprehension per chart type, check labeling against human annotations, and run case studies on public visualizations and a redesign.

arxiv.orgpaperAdded 27 Oct 2026, published 25 Oct 2026🧠 AI

InteractiveWorld is an environment for closed-loop interaction between an AI assistant and a simulated human agent who sees the scene, follows instructions and acts independently. InteractiveBench, built on it, has 120 multi-turn tasks in six categories scored by verifiable, state-based rewards. Across 12 frontier MLLMs, the best model, Claude Fable 5.1, reaches at most 20% on long-horizon Guidance and 25% on Planning tasks.

arxiv.orgpaperAdded 27 Oct 2026, published 25 Oct 2026🧠 AI

DRILS is an LLM-based agentic framework for requirements change impact analysis. It uses multiple agents to generate candidate impact sets, with an LLM-as-a-judge promoting diversity in selection. It adds RE-RAG, a tuned retrieval technique, and a requirement dependency graph as a knowledge source. On three curated datasets it reaches 87.1% recall, 76.3% precision and 81% F1 at lower cost than baselines. The authors release the datasets.

arxiv.orgpaperAdded 27 Oct 2026, published 24 Oct 2026🧠 AI

AutoTavern is a self-improving system of specialised LLM research agents with a typed shared memory and verification gates. In 20 days it built TavernGym, a full-content Hearthstone Battlegrounds simulator, and 23 versions of TavernAgent, a PPO-trained player. TavernAgent averaged placement 1.60 with 80% first places over its last 20 live games, beating the top ten players (2.34, 59%). On TavernBench (960 tasks), it answered 91% of play tasks against at most 63% for frontier LLMs.

arxiv.orgpaperAdded 27 Oct 2026, published 24 Oct 2026🧠 AI

TriageRAG is a failure-aware self-training method for adapting retrieval-augmented LLMs to specialized domains using only an unlabeled corpus. It probes the student on self-generated questions with and without retrieved context, sorts them into known, solved, context-failure and retrieval-failure, and routes each to a matching training signal. With Qwen3-4B on 11 datasets it beats budget-matched SimRAG by 1.2-6.4 points and recovers closed-book accuracy where retrieval hurts.

arxiv.orgpaperAdded 27 Oct 2026, published 24 Oct 2026🧠 AI

RALE (Representation-Aware Logical Ensemble) is a method for detecting vulnerabilities in C source code with LLMs trained on different code representations: raw source, Abstract Syntax Trees and Control Flow Graphs. The authors show the representations suit different vulnerability classes, and combine the models with explicit logical rules and majority voting. Single-representation models reach up to 73% F1, while the ensemble reaches 78%.

arxiv.orgpaperAdded 28 Oct 2026, published 24 Oct 2026🛡️ Security

RALE (Representation-Aware Logical Ensemble) is a method for LLM-based vulnerability detection in C code. The authors train LLMs on raw source code, Abstract Syntax Trees and Control Flow Graphs, and find each representation works best for certain vulnerability classes. Combining the models with explicit logical rules and majority voting raises F1 from at most 73% for a single model to 78%, with gains in accuracy, precision and recall.

arxiv.orgpaperAdded 28 Oct 2026, published 24 Oct 2026🧠 AI

VulDec is a dataset and framework for detecting vulnerabilities in binaries by first decompiling them with fine-tuned LLMs. The dataset aligns real-world Assembly-C function pairs across several architectures with CWE labels. Tuned neural decompilers reach about 56% Edit Similarity and 43% BLEU-4, and tuned detectors reach a macro F1 of 0.61 on a realistic test set with a 3% vulnerability rate.

arxiv.orgpaperAdded 28 Oct 2026, published 24 Oct 2026🧠 AI

Provable Security of SHA-256-Based LMS Revisited in the Ideal-Cipher Model is an October 2026 ePrint paper by Chun Guo, Hao Cheng, Bart Preneel and Meiqin Wang. It shows forgery attacks on LMS_SHA256_M32 and LMS_SHA256_M24 that exploit Merkle-Damgård structure, so the Katz and Fluhrer random-oracle bounds do not apply. It proves ideal-cipher-model bounds: about 256 minus log2 of the block count for M32, and a minimum of two terms for M24.

eprint.iacr.orgpaperAdded 26 Oct 2026, published 24 Oct 2026🛡️ Security

The paper builds two corpora, EventTyp and the human-validated EventTypGold, to study how people infer typicality and atypicality of events from what is said. Large-scale human and LLM experiments show that three CUES predict these inferences in humans: Commonsense Knowledge, Utterance Emphasis and Snark. In most settings LLMs lack the commonsense knowledge to judge an event's general typicality, so they struggle to follow the CUES.

arxiv.orgpaperAdded 27 Oct 2026, published 23 Oct 2026🧠 AI

NTBench (Near-Target Benchmark) tests LLM and agent robustness to near-target prompt injection, where injected instructions steer a response toward a plausible but incorrect or harmful outcome instead of abandoning the task. It builds reference-guided target-instruction pairs on existing datasets, with human validation, and evaluates seven LLMs on ten language datasets and two agent benchmarks. Attack success rose 9.3 to 35.8 points (incorrect targets) and 1.7 to 17.3 points (harmful targets) over fixed-target baselines.

arxiv.orgpaperAdded 26 Oct 2026, published 23 Oct 2026🛡️ Security

BiLoCo is a post-training quantization method for FP4 LLMs that adds low-rank corrections whose left and right factors are stored as binary ±1 sign vectors instead of BF16. This uses 16x less storage per rank-one component and lets the correction run on CUDA cores alongside the FP4 tensor-core GEMM. On 4B to 72B LLMs it matches or improves memory-matched SVDQuant. A fused B300 kernel is at most 11% slower than FP4-only GEMM at batch size 1 for ranks up to 512.

arxiv.orgpaperAdded 26 Oct 2026, published 23 Oct 2026🧠 AI

This paper studies reverse sycophancy: whether users bend toward a confident, consistent, wrong LLM. In a within-subjects experiment, 32 participants learned science topics from five GPT-5-mini assistants with 0 to 100% of facts replaced by falsehoods. Scores fell from 7.4 to 2.9 out of 10, planted falsehoods were chosen on 60% of falsified items against 11% of correctly taught ones, and confidence stopped tracking correctness. It introduces a baseline-corrected Human Deference Index.

arxiv.orgpaperAdded 24 Oct 2026, published 22 Oct 2026🧠 AI

TTR-Compressor is a sequence compression method that treats input tokens as a test-time regression problem. The regression is learned end-to-end by backpropagation, and its coefficients are solved in closed form in the forward pass to give the compressed representation. As a block-wise sparse attention compressor it improves RULER over attention-based compressors by a large margin. As a visual token merger for multimodal LLMs it beats the MLP merger and matches it with 55.6% fewer visual tokens.

arxiv.orgpaperAdded 24 Oct 2026, published 22 Oct 2026🧠 AI

This empirical study benchmarks recent LLMs against fine-tuned encoder-decoder models on knowledge graph-to-text generation, using reference-based metrics and graph-grounded, reference-free LLM-based evaluation validated against human annotations. Recent LLMs narrow or close the gap on reference-based metrics and do better on graph-grounded ones. Prompted models make fewer contradictions and triple omissions, and fine-tuned models' errors grow faster with graph size.

arxiv.orgpaperAdded 24 Oct 2026, published 22 Oct 2026🧠 AI

The Geometry of Metaphor is an October 2026 arXiv preprint that uses mechanistic interpretability to find causal subspaces for metaphor in LLMs of different architectures and sizes. The authors steer output on classification tasks across parts of speech. They also steer generation tasks, showing the metaphoricity of model output can be controlled. The abstract gives no model names or metrics.

arxiv.orgpaperAdded 24 Oct 2026, published 22 Oct 2026🧠 AI

DuoCook is an Overcooked-inspired benchmark in which one vision-language model controls two cooks from top-down views in Unreal Engine environments built with coding agents. It has 20 maps in five categories covering equipment contention, changing terrain and handoff dependencies. Four VLMs were compared; the best, gpt-6-astra, reached 26.2% mean order completion. A notebook memory ablation had mixed effects: positive for Claude and Qwen, negative for GPT and Gemini.

arxiv.orgpaperAdded 24 Oct 2026, published 22 Oct 2026🧠 AI

This paper trains small open language models with reinforcement learning to select tools for multi-step instructions over corpora of more than ten thousand tools. A Qwen3-8B model post-trained with the recipe outperforms Llama-4-Maverick (402B) on ToolBench. The trained model transfers to different retrievers and generalizes to new tool corpora in other domains, without relying on proprietary LLMs like Claude or ChatGPT.

arxiv.orgpaperAdded 24 Oct 2026, published 22 Oct 2026🧠 AI

This paper studies evidence-grounded health summarization for menstrual health monitoring. It combines structured evidence conditioning with RAG in a prompt-based constrained generation framework, and scores 4,500 summaries from three LLMs with NLI, atomic factuality metrics and readability measures. Grounding constraints improved uncertainty preservation and generally kept or improved faithfulness against retrieved evidence, while omission effects varied by model. Clinical alignment came with harder reading and more medical terminology.

arxiv.orgpaperAdded 24 Oct 2026, published 22 Oct 2026🧠 AI

Rate-Controlled Refresh (RCR) is a KV cache re-selection method for multimodal LLMs in long-form generation. Attention drift decides when and where to re-select visual tokens, and a per-sample controller sets how often. RCR beats InfiniGen by 1.6-3.8 ROUGE-L points on five of six model-task settings and stays within 1.7 points of exact per-step re-selection. On video description, where drift is homogeneous, it no longer beats a fixed threshold.

arxiv.orgpaperAdded 24 Oct 2026, published 22 Oct 2026🧠 AI

SecLog-Bench is a benchmark for LLM reasoning over cybersecurity logs, built on an extensible simulation platform covering 29 attack scenarios. It combines audit logs, application logs and network traffic, with attack-stage labels and step-by-step evidence chains. Eight LLMs were tested on post-detection attack-progress reasoning: the best reached 44.44% accuracy, 45.07% Macro-F1 and 54.87% Stage Reasoning F1.

arxiv.orgpaperAdded 23 Oct 2026, published 22 Oct 2026🛡️ Security

SecLog-Bench is a benchmark for LLM reasoning over cybersecurity logs. A simulation platform with 29 attack scenarios generates audit logs, application logs and network traffic, with attack-stage labels and step-by-step evidence chains. Eight mainstream LLMs were tested on attack-progress reasoning: the best reached 44.44% accuracy, 45.07% Macro-F1 and 54.87% Stage Reasoning F1. Correct predictions often came with flawed reasoning.

arxiv.orgpaperAdded 23 Oct 2026, published 22 Oct 2026🧠 AI

The SC-matrix is an evaluation framework for multilingual LLM sycophancy. It crosses a model's competence (correct answer with no pressure) with its sycophancy (keeping that answer under pressure), so agreeing with a false premise can be told apart from not knowing the answer in that language. Results: equal aggregate sycophancy rates hide different failure mixes, high competence does not always protect, and reasoning-oriented mitigations beat directive ones.

arxiv.orgpaperAdded 24 Oct 2026, published 22 Oct 2026🧠 AI

Paper on memory injection, where prompt injection persists across sessions through the long-term memory of LLM-as-a-service platforms. It presents a general threat model and shows all NIST attack goals are achievable on ChatGPT Instant 5.3, Claude Sonnet 4.6 and Gemini 3 Flash. Vectors include personal memories, projects, skills, organization instructions and MCP resources. On Claude, shared memory lets an attacker compromise multiple users in an organization and exfiltrate prompts. Nearly half of the derived mitigations are unimplemented.

arxiv.orgpaperAdded 24 Oct 2026, published 20 Oct 2026🛡️ Security

InterWebCoder is a framework for interactive webpage code generation by multimodal LLMs. It synthesizes self-contained webpages with executable test scripts from recorded interaction trajectories, repairs failures using runtime evidence, then trains with test-ID-weighted SFT and GRPO using test pass rates and visual similarity as rewards. The pipeline yields 1,736 verified samples from 2,058 tasks. A 9B model scores 48.54 on IWR-Bench and 48.42 on Vision2Web, and reward latency drops over 97%.

arxiv.orgpaperAdded 22 Oct 2026, published 20 Oct 2026🧠 AI

DAR-MGSM (Dialectal Arabic Multilingual Grade School Math) is a manually curated benchmark for mathematical reasoning in Dialectal Arabic. It covers five major Arabic dialects plus Modern Standard Arabic, in a translated-only version and a translated and culturally localized version. Eleven Arabic-centric and multilingual LLMs were tested: accuracy generally drops on the localized version, and a mechanistic interpretability analysis shows the two variants induce distinct internal reasoning trajectories.

arxiv.orgpaperAdded 22 Oct 2026, published 20 Oct 2026🧠 AI

Code2Video-Compass is a benchmark of 2,104 tasks for code-based video generation, where multimodal LLMs write executable programs that produce complete videos from text (T2V), image (I2V) or video (V2V) inputs. Tasks span 10 categories and 24 subtypes. Scoring uses 12 objective metrics and five MLLM-judge metrics. Across 15 proprietary and open-weight models from nine families, performance falls as construction, dynamic animation and state-dependency complexity rise.

arxiv.orgpaperAdded 22 Oct 2026, published 20 Oct 2026🧠 AI

DeclineCraft is a dynamic RAG abstention benchmark whose unanswerable questions target Minecraft feature additions made after a corpus cut-off, with audits for parametric exposure. Across 27 LLMs, single-shot retrieval raises Correct Abstention Rate in 38 of 40 configurations, while iterative RAG lowers it in 30 of 40. Enabling thinking without retrieval lowers abstention for 8 of 13 models, shifting responses to premise denial.

arxiv.orgpaperAdded 22 Oct 2026, published 20 Oct 2026🧠 AI

OpenVLHarness is a multimodal reasoning harness for MLLMs built on a capability layer (task-level tool interfaces plus evidence rendering) and persistent multimodal memory that stores images, structured data and web references as addressable artifacts. A coding interface runs generated programs over stored data. Across 22 benchmarks and eight backbones it adds 9.7-12.6 points over no-tool baselines, and GPT-6 Luna at medium effort beats Codex and Visual Sketchpad at maximum effort using at most 37% of their tokens.

arxiv.orgpaperAdded 21 Oct 2026, published 19 Oct 2026🧠 AI

The paper introduces PhysRSI-eval, a five-axis metric for physical recursive self-improvement: first-attempt performance, efficiency, success, budget and improvement from failure. Several VLMs are tested on 18 manipulation tasks under a fixed interaction budget. It proposes the Physical Barrel Hypothesis and studies observational scaling. Workspace interventions show that removing history lowers late-round success, while frozen workspaces still help on unseen layouts.

arxiv.orgpaperAdded 21 Oct 2026, published 19 Oct 2026🧠 AI

TokenGate is a small drop-in component that adds end-to-end learnable token pruning to trained LLMs for long-context prefill. It uses token-gated attention, inspired by Mixture-of-Experts, to re-weight attention logits with token importance scores, which gets gradients through the non-differentiable top-K step. The authors report 1.4-11.3x faster time-to-first token and 1.5-3.4x lower peak memory with no accuracy loss, on Qwen3 and Qwen3.5.

arxiv.orgpaperAdded 21 Oct 2026, published 19 Oct 2026🧠 AI

The paper is a large-scale study of 37 security skills run through Claude Code with four LLMs on 100 real Linux kernel vulnerabilities, 14,800 runs in all. Only 24.4 percent of runs found the correct function and 15.7 percent classified the vulnerability correctly. It proposes SKILL-SIF, which evolves skills from failure patterns and raised fully correct detection on 100 unseen vulnerabilities from 2 to 16 percent.

arxiv.orgpaperAdded 21 Oct 2026, published 19 Oct 2026🧠 AI

Empirical study of 37 security skills run through Claude Code with four LLMs on 100 real Linux kernel vulnerabilities across ten classes (14,800 runs). 24.4% of runs found the correct function, 18.3% the function and statement, and 15.7% the vulnerability class. The authors propose SKILL-SIF, which evolves skills from failures and raised fully correct detection on 100 unseen vulnerabilities from 2% to 16%.

arxiv.orgpaperAdded 21 Oct 2026, published 19 Oct 2026🛡️ Security

OwlSight is a chain-of-thought monitor that uses "Factor", a monitoring requirement written into the prompt, so one monitor can cover many tasks across capability, safety and trustworthiness. Factors come at fine granularity (a specific leaked email address) and coarse granularity (privacy leakage). On 15,971 reasoning trajectories it beats general-purpose LLMs by 1.32-30.57 points (fine) and 11.71-27.81 points (coarse), and reaches state-of-the-art on OpenAI's CoT monitorability benchmark. It generalizes to unseen models and resists Factor rephrasing, CoT compression and obfuscation.

arxiv.orgpaperAdded 20 Oct 2026, published 16 Oct 2026🧠 AI

This paper builds a benchmark of 500 debatable questions across ten domains to measure how user narratives steer LLM agreement. Each question has 11 matched conditions (one neutral, five mirrored narrative levels per opposing stance), giving 5,500 instances across seven LLMs. Causal framing and worldview synthesis raise agreement, personal experience can lower it, and GPT-5.1 ablations show bullet formatting also increases agreement.

arxiv.orgpaperAdded 20 Oct 2026, published 16 Oct 2026🧠 AI

LexPilot is a benchmark linking proactive legal interaction to claim-level verdict prediction, built from 432 real Chinese civil judgments across 36 causes of action. It pairs expert-reviewed complete and underspecified consultation histories with court verdicts. Current LLMs score on average 23.84% lower on judgment prediction in underspecified consultations than in fully specified ones, and the paper analyses the bottlenecks.

arxiv.orgpaperAdded 20 Oct 2026, published 16 Oct 2026🧠 AI

EDIC is a decomposed reward framework that trains LLMs to decide, for each instruction embedded in external data such as documents or tool returns, whether to follow or reject it. A judge initialized with SFT and refined with GRPO attributes compliance per instruction and yields separate rewards for beneficial and harmful compliance. On Qwen3-4B it raises harmful-rejection by up to 7.7% and Arena-Hard-Auto utility by 12.5% over the strongest prior defense.

arxiv.orgpaperAdded 20 Oct 2026, published 16 Oct 2026🧠 AI

EDIC is a decomposed reward framework for per-instruction compliance in LLMs that read external data, where one document or tool return mixes legitimate directives with injected ones. A generative judge, trained with SFT then GRPO, attributes compliance per instruction and produces RL rewards for beneficial and harmful compliance. On Qwen3-4B it raises harmful rejection by up to 7.7% and Arena-Hard-Auto utility by 12.5% across five benchmarks.

arxiv.orgpaperAdded 20 Oct 2026, published 16 Oct 2026🛡️ Security

R^2ECAP (Reused Reasoning-Cache Poisoning) is a black-box attack that plants poisoned reasoning traces through a chat interface, exploiting LLM systems that reuse KV caches of reasoning. A preceding trace steers the model, and the retained trace keeps residual adversarial influence in its cache after the preceding trace is discarded. On three open-weight LLMs and three datasets it reaches up to 79% ASR, beats the white-box HijackKV in 8 of 9 settings, and reaches 46% ASR in a reasoning-trace retrieval system. Only 1% of retained responses are flagged by a popular prompt injection detector.

arxiv.orgpaperAdded 20 Oct 2026, published 16 Oct 2026🛡️ Security

Doc2Harness is a two-stage benchmark for harness engineering. In Stage 1 a build model, working inside a fixed builder harness, gets an interface-only contract skeleton plus a development document consolidating behavior from twelve frontier harnesses, and builds a full candidate harness in an isolated container. In Stage 2 the frozen harness runs in separate per-test-case containers, driven by a scripted mock model, and is scored on four channels: requests sent to the model, files on disk, exit code and command output. Across 19 frontier models with three builds each on 328 test cases, GPT-6 Astra leads at 53.05% against 27.44% for the weakest, and failures cluster around how cross-cutting functionality is split across modules.

huggingface.copaperAdded 16 Oct 2026🧠 AI

Introduces Librarian Agents, where a builder agent turns raw unstructured documents into a hierarchical filesystem of synthetic files and reference links, forming a workspace for a downstream answering agent. Workspace construction is cast as a Markov decision process and the builder policy is trained with reinforcement learning to maximize downstream answering accuracy. It beats the strongest retrieval and precomputation baselines by 33 points on BrowseComp-Plus, amortizes construction cost over repeated queries, and generalizes zero-shot to BRIGHT with a 31-point average gain.

arxiv.orgpaperAdded 17 Oct 2026, published 15 Oct 2026🧠 AI

DropKV is a KV-cache eviction framework for long-context LLMs. It decouples the joint eviction decision into independent per-token scores based on residual-output perturbation, which avoids the NP-hard combinatorial problem. It has a provable constant-factor approximation guarantee under a cone condition, which the authors verify empirically on three long-context LLMs. The fused Triton kernels never materialize the full attention matrix. They give a 19.8x scoring-kernel speedup and up to 9.9% end-to-end prefill speedup. On RULER and LongBench it beats state-of-the-art eviction baselines at equal cache budgets.

arxiv.orgpaperAdded 17 Oct 2026, published 15 Oct 2026🧠 AI

GraphACM merges equivalent states from multiple rollouts of the same LLM agent task into a directed, action-labeled graph, so actions can be compared at shared decision states. It defines node-level and global action-consistency measures, and uses the global score as an uncertainty proxy to predict task success or failure. Across three LLMs on ALFWorld, AppWorld and WebShop it reaches an average AUROC of 0.811, above the strongest behavioral-consistency baseline (0.729) and verbalized confidence (0.769). It also shows where in a trajectory agent actions diverge.

arxiv.orgpaperAdded 17 Oct 2026, published 15 Oct 2026🧠 AI

GACC-Bench is a benchmark for Goal-Adaptive Context Compression, testing how well LLMs retain goal-relevant information when compressing a source context under a length budget before downstream questions are known. It has 500 goals and 2,571 human-reviewed questions over 281 document and agent-trace contexts from seven collections; a fixed reader answers the withheld questions from the compressed context. Ten LLM compressors show large gaps to full-context performance, especially in answering every question tied to a goal. Target-goal conditioning improves accuracy over goal-agnostic compression, while mismatched goals hurt it.

arxiv.orgpaperAdded 17 Oct 2026, published 15 Oct 2026🧠 AI

INFUSE is an end-to-end benchmark for indirect prompt injection (IPI) across long-horizon agent workflows (shell commands, code changes, dependency installs, remote services). It has 40 realistic tasks and a taxonomy of 56 attack scenarios in five threat categories, with attacks injected via interaction surfaces encountered mid-task. Across 19 LLMs (GPT, Claude, Gemini, Llama, Gemma, Qwen) and 41,824 adversarial trajectories, mean attack success was 19.6% (LLM-judged) and 15.4% (heuristic), per-model 3.2–41.9%; guard models gave limited protection, while prompt-level hardening cut mean ASR from 58.2% to 14.3%.

arxiv.orgpaperAdded 17 Oct 2026, published 14 Oct 2026🛡️ Security

Draft ERC from Quantova Inc that has Ethereum bridges accept cross-chain message batches only when a committee signs the Merkle root with RFC 8554 LMS hash-based signatures. The guard contract requires a strict majority of distinct unfrozen keys, records used leaves, freezes any key that signs a leaf twice, binds digests to domain tag, chain ID, guard address and epoch, and enforces a 60-second delay. It reports gas benchmarks (about 205k per LMS W4 height-10 verify, 12.44M worst-case rotation under EIP-7825), a hybrid LMS+ECDSA migration mode, and a NioLabs assessment with no critical or high findings. It argues that Shor's algorithm makes secp256k1 and BLS committee keys a quantum theft target.

ethereum-magicians.orgpaperPublished and added 14 Oct 2026🛡️ Security

Proposes mmVB, a lightweight plug-and-play module (1.6M trainable parameters) that assigns query-conditioned per-image visual token budgets using cross-image context for multimodal LLMs, without retraining the base model or accessing attention matrices (so it is compatible with FlashAttention). It cuts visual input tokens 9x, giving 5x faster time-to-first-token and 7x fewer prefill FLOPs while keeping 94.4% of full-resolution accuracy across eight multi-image reasoning benchmarks. It matters because existing token pruning, merging and compression methods target single-image redundancy and either need attention maps or retraining.

arxiv.orgpaperAdded 14 Oct 2026, published 13 Oct 2026🧠 AI

Plan-Bench is a benchmark of 699 curated samples across 16 subclasses and three dimensions. It tests whether vision-language models can reason over spatial and temporal information and produce multi-step plans under constraints for embodied tasks, with action execution isolated out. Across 28 VLMs the best model reaches only 75.0% accuracy. Scaling generally helps, and larger models look more promising as high-level planners. Code and benchmark are released.

arxiv.orgpaperAdded 14 Oct 2026, published 13 Oct 2026🧠 AI

Game2Real-Bench tests whether vision-language model performance on procedurally generated toy games predicts real-world visual reasoning, and whether game examples help transfer. It has 1,000 verifiable game samples and 400 real-world samples, matched by shared reasoning requirements (fine-grained perception, spatial reasoning, constraint reasoning, sequential planning). Across 25 VLMs, game and real scores correlate strongly (Pearson/Spearman 0.935). One-shot game-to-real transfer on four models improves constraint reasoning for several but hurts fine-grained perception. The authors also release Game2Real-Dataset, with about 24K verified game samples across 18 tasks and 65 QA families.

arxiv.orgpaperAdded 14 Oct 2026, published 13 Oct 2026🧠 AI

OmniCompass is a manually annotated benchmark for video grounding and understanding with two tracks: OmniCompass-VL (820 videos, 2,460 instances; vision-language models) and OmniCompass-AV (240 videos, 720 instances; joint audio-visual evidence for omni-modal models). Each track covers temporal grounding, spatial grounding and QA across short/medium/long videos, and each instance is run in paired text-only and text+image target specifications. Evaluating 50 MLLMs shows open-source models lag proprietary ones, especially on audio-visual understanding, that precise temporal/spatial grounding remains hard, and that predictions are inconsistent across the paired specifications.

arxiv.orgpaperAdded 14 Oct 2026, published 13 Oct 2026🧠 AI

EmoGenie is a benchmark for emotions expressed and evoked by text-to-image outputs. It has 6,400 images from eight T2I models using 800 emotion-targeted prompts. It includes EmoGenie-QA (90,000 closed-set and 1,057 open-ended questions) and EmoGenie-Assess (13,175 mean opinion scores for valence, arousal, dominance and aesthetic quality). The authors also propose EmoProbe, which pairs emotion QA with response-aligned, dimension-specific regression heads. They report that it beats baselines on emotion QA and correlates better with human ratings. The work shows how reliably MLLMs interpret affect in generated images.

arxiv.orgpaperAdded 14 Oct 2026, published 13 Oct 2026🧠 AI

The paper studies regressive sycophancy under fabricated evidence: a medical VLM gives a correct answer, then abandons it when the user pushes back with a fake prescription. It tests four open VLMs on 400 VQA-RAD yes/no questions in 19,200 multi-turn conversations. Without a synthetic-document disclosure, models abandon 80.1% of 6,488 correct answers, and the same prescription raises Qwen2.5-VL-7B's flip rate from 45.9% to 89.6% while lowering Qwen3-VL-8B's from 77.7% to 50.9%. Models often reject the prescription yet keep the correct answer in only some conversations (915 of 3,232), and 40% of those flip later. System-prompt disclosure of synthetic documents cuts one model's flip rate from 77.8% to 47.3%. The authors conclude that recognising a forgery does not stop a model from adopting the forged answer.

arxiv.orgpaperAdded 14 Oct 2026, published 13 Oct 2026🧠 AI

SecReason is an in-stream defense for reasoning LLMs that write code. Reasoning models draft code early in their streamed chain of thought, so SecReason scans that stream with dual-track incremental analysis: security rules are checked on each emitted statement, and dataflows are traced across the accumulated code. When a rule fires, it interrupts decoding, inserts a warning into the stream and resumes, so the model fixes its draft within the same reasoning trace instead of regenerating the full response. On CWEval and SecCodeBench, runs delivering secure and functionally correct code rise from below 15% to 72.8%–98.1%. It uses fewer tokens than post-hoc repair and limits functional regressions when rules misfire on secure code.

arxiv.orgpaperAdded 15 Oct 2026, published 10 Oct 2026🧠 AI

SecReason is an in-stream defense for reasoning LLMs that monitors code drafted inside the streamed chain of thought. Dual-track incremental scanning applies security rules per statement and traces dataflows across accumulated code; when a rule fires, generation is interrupted, a warning is inserted, and decoding resumes so the model repairs the draft in the same trace. On CWEval and SecCodeBench it raises the share of secure and functionally correct runs from below 15% to 72.8%–98.1%, uses fewer tokens than post-hoc repair, and limits regressions from false warnings.

arxiv.orgpaperAdded 16 Oct 2026, published 10 Oct 2026🛡️ Security

SecReason is an in-stream defense for reasoning LLMs that monitors code drafted inside the streamed chain of thought. Dual-track incremental scanning applies security rules per statement and traces dataflow across accumulated code. When a rule fires, generation is interrupted, a warning is inserted and decoding resumes, so the model repairs its draft without regenerating the full response. On CWEval and SecCodeBench it raises the share of secure and functionally correct runs from below 15% to 72.8%–98.1%, uses fewer tokens than post-hoc repair, and limits functional regressions from false warnings.

arxiv.orgpaperAdded 15 Oct 2026, published 10 Oct 2026🛡️ Security

SecReason is an in-stream defense for reasoning LLMs that write code. Reasoning models draft code early in their streamed chain of thought, and SecReason scans that draft incrementally with security rules, checking each statement as it is emitted and tracing dataflows across the accumulated code. When a rule fires, it interrupts generation, inserts a warning into the stream, and resumes decoding, so the model repairs the draft within the same reasoning trace. On CWEval and SecCodeBench it raises the share of runs that deliver secure and functionally correct code from under 15% to 72.8%–98.1%. It uses fewer tokens than post-hoc repair and limits functional regressions when rules misfire on secure code.

arxiv.orgpaperAdded 16 Oct 2026, published 10 Oct 2026🧠 AI

Presents an EM side-channel attack on LLaMA-3.2-1B-Instruct running on a Raspberry Pi 5, showing passive physical measurements leak generated tokens and user prompts. Three attacks: a profiling attack (97.2% accuracy over 50 medical/privacy keywords), a hidden-state reconstruction attack predicting tokens from the 128k vocabulary (94.3% on profiled classes; 67% of unseen words ranked in top 100), and a hidden-state inversion attack reconstructing prompts semantically. Shows that local edge inference does not guarantee confidentiality of tokens and inputs.

arxiv.orgpaperAdded 16 Oct 2026, published 9 Oct 2026🛡️ Security

Introduces EICS (Evaluation-Invariant Concept Stability), a metric for checking whether representations consistently encode a concept under controlled one-concept interventions. Also introduces CASP-VL, a benchmark of causally controlled spatial perturbations derived from Visual Genome scene graphs. Across several VLM architectures, spatial relations are represented much less faithfully than lexical information, even for models well above chance on compositional benchmarks. The relation is identifiable at the predicate's own tokens but is lost when the sentence is aggregated into a global representation (EOS token or mean pooling), a failure that output-based evaluation does not reveal.

arxiv.orgpaperAdded 12 Oct 2026, published 9 Oct 2026🧠 AI

DualQuant is a calibration-free post-training quantization framework for LLM weights and activations. It minimizes weight-reconstruction MSE using variable splitting and block coordinate descent with closed-form subproblem updates. The resulting scaling factors double as per-channel activation rescalings, so no calibration data is needed. It supports the MX family and NVFP4 block formats. On Llama and Qwen models it matches or beats calibration-based SmoothQuant on WikiText-2, C4 and zero-shot tasks. A real packed W4A4 path for Llama-3.1-8B on an RTX 4090 gives up to 2.61x memory reduction and 2.46x throughput over BF16.

arxiv.orgpaperAdded 12 Oct 2026, published 9 Oct 2026🧠 AI

Proposes MASTIF (Multi-Agent System TestIng Framework) for evaluating Agentic Web stacks. It reports a baseline assessment of 3 LLMs, 3 protocols and 4 agent implementation frameworks on 300 Mind2Web benchmark tasks. Results suggest that choosing a task-specific stack can improve task completion rate, efficiency and social value stability, and can reduce verbosity and completion time. This gives developers reference metrics for choosing among fast-moving agent protocols and frameworks.

arxiv.orgpaperAdded 12 Oct 2026, published 9 Oct 2026🧠 AI

Studies curriculum design for GRPO by re-estimating sample difficulty under the current policy and using a staged curriculum that progressively shifts the active difficulty band toward harder samples. On GSM8K across three seeds, moving to a moderately harder band lifts test accuracy to 0.6591 ± 0.0077 versus 0.6449 ± 0.0024 for fixed-medium selection, while going to a still harder band drops it to 0.6405 ± 0.0053. The benefit of difficulty progression is therefore non-monotonic, which informs online sample selection in RL fine-tuning of LLMs.

arxiv.orgpaperAdded 12 Oct 2026, published 9 Oct 2026🧠 AI

SWE-DeceptionBench is a benchmark of 720 Chinese and English test instances measuring LLM deception across six software development lifecycle stages. It separates Active Deception (the model recognizes infeasible or conflicting requirements but presents them as achievable) from Passive Deception (the model helps users mislead others), and it tests Neutral, Reward and Pressure conditions. Across seven mainstream LLMs, Neutral-condition deception reached 49.88% (Active) and 64.29% (Passive). Guard models missed many deceptive requests, and Self-Reminder reduced but did not eliminate deceptive responses.

arxiv.orgpaperAdded 12 Oct 2026, published 9 Oct 2026🧠 AI

First systematic empirical study of Cross-chain Verification Vulnerabilities (CVVs) in bridge smart contracts, such as flawed validation of state proofs, message sender identities and signatures. From 588 audit reports across 36 bridge projects it derives 11 CVV patterns in 5 categories and builds a dataset of 830 instances across 786 contract files, mapped to reports, source versions and pattern labels. Existing tools cover few CVV patterns, and only two of four tested LLMs beat a 29.3% Macro-F1 constant baseline (best 51.2%), which motivates detectors built for cross-chain verification semantics.

arxiv.orgpaperAdded 12 Oct 2026, published 9 Oct 2026🛡️ Security

DecepEval is a benchmark of 1,532 instances across 3 task families and 28 professional scenarios designed to measure deception in LLM agents, pairing neutral and induced versions of each task to isolate condition-dependent changes in deception rates. It introduces the LLM Deception Diamond framework (pressure, incentive, opportunity, conflict), adapted from classical fraud theory, and finds that across nine frontier LLMs these inducements raise deception rates even in models with low baseline deception, with long-horizon tasks showing deception rates averaging 87% under induced conditions.

huggingface.copaperAdded 8 Oct 2026🧠 AI

OmniCapBench is a benchmark of 786 densely annotated videos that evaluates audio-visual captioning in multimodal LLMs. Instead of scoring free-form captions, it asks models for sets of atomic, verifiable units across three tracks (entity references, visual shots, audio events). These are scored with deterministic constraint checks plus localized LLM-based semantic comparisons, which gives coverage, localization and stable scoring together. Tests on frontier MLLMs show strong local perception but weak long-horizon audio-visual reasoning, especially identity drift and cross-modal misalignment.

arxiv.orgpaperAdded 9 Oct 2026, published 8 Oct 2026🧠 AI

WOVEN is a training source and benchmark for visual transition reasoning in multimodal LLMs. It has 36,076 examples spanning 20 scene types, 5 action types and 8 reasoning types, built from rollouts of video-pretrained generative models. Evaluating 38 frontier MLLMs (including GPT-5.4 and Qwen3-VL-235B-A22B) shows systematic deficits that persist with scale. Training on WOVEN subsets of about 2,000 items improves 22 of 26 external benchmarks by up to 27.3 points, and WOVEN data can replace 30-50% of a task's own training data. The authors derive a recipe: select supervision by the reasoning operation it teaches rather than by scene or action, and prefer larger visual-state changes.

arxiv.orgpaperAdded 9 Oct 2026, published 8 Oct 2026🧠 AI

ORCAGen uses RAG over a curated knowledge base of malware procedures and active-defense strategies, plus structured prompting, to have LLMs generate proof-of-concept malware and matching deception orchestration code offline. Strategies are validated against the PoC malware before being added to a runtime playbook that enforces only the verified, deterministic logic. It is evaluated on GPT-4o, GPT-5.5, Gemini 3.5 Flash, Qwen3-Coder and Claude Sonnet 4.5 over synthesized scenarios and 150 real keylogger, infostealer and ransomware samples, measuring execution success, hallucination rate, refinement effort, deception effectiveness and runtime overhead.

arxiv.orgpaperAdded 9 Oct 2026, published 8 Oct 2026🛡️ Security

Introduces AgenticBBO-Bench, a benchmark for LLM-agent black-box optimization across synthetic functions, hyperparameter optimization, database tuning, chip design and molecular design under a unified finite-budget protocol. Agentic BBO beats direct LLM methods in all five domains and the best numerical optimizers in four. Ablations show extra numerical tools don't consistently help, task semantics help more than specific priors, and numerical optimizers can absorb gains from agent-found trajectories. A five-task frontier challenge evaluates seven LLMs under the Codex agent harness, with GPT-6 Astra and DeepSeek-V4.1-Flash on the performance-cost Pareto frontier.

arxiv.orgpaperAdded 9 Oct 2026, published 8 Oct 2026🧠 AI

Paper measuring prefix-based training-data extraction in instruction-tuned open-weight LLMs under an empty prompt versus retrieved contexts of varying relevance (RAG), using paired item-level suffix scoring across three models. It finds extractable memorization has a context-robust core and a context-sensitive boundary: context suppresses some exposures but also enables new ones missed by prefix-only evaluation. This qualifies the claim that RAG reduces memorization risk and matters for threat-modeling deployed LLM systems.

arxiv.orgpaperAdded 9 Oct 2026, published 8 Oct 2026🛡️ Security

Evaluates Jev, a "System One" decision model that returns a choice among given options instead of generating text, on 13 multiple-choice benchmarks covering knowledge, reasoning and multilingual understanding. It is compared with 19 LLMs in frontier, representative and small tiers. Jev is competitive with frontier LLMs on knowledge and commonsense tasks, and best on MMLU-Redux and ARC-Challenge. It beats most representative LLMs and all small ones outside math, but is 17.7 points below the frontier median on MathQA and scores lower than all 19 LLMs there. This suggests specialized decision models can match general LLMs on knowledge-driven decisions but not on multi-step calculation.

arxiv.orgpaperAdded 9 Oct 2026, published 8 Oct 2026🧠 AI

Mazor, Morgan and Pass prove that LLMs, modeled as next-token predictors, are equivalent to monotone (order-preserving) compression algorithms: either can be built from the other with the same error up to an additive gap of 2. They show that monotonicity is needed for this equivalence if and only if infinitely-often one-way functions exist. As a corollary, next-bit pseudoentropy of a distribution is equivalent to monotone incompressibility. This gives a theoretical basis for empirical findings that compression ability tracks LLM benchmark performance.

eprint.iacr.orgpaperAdded 11 Oct 2026, published 8 Oct 2026🧠 AI

RunningTab is a framework that gives LLM agents performing 'direct workspace interaction' (searching and reading files from a terminal without indexing to produce a deliverable) an environment-side 'tab': a per-task record, maintained outside the model's context, of requirements the agent has declared, excerpts read with provenance, and listed-but-unopened files flagged as candidates. The agent can check each requirement against matching excerpts or candidates, mark it resolved or deferred with a reason, and is forced through a finish check if requirements remain open when it tries to complete the task. The authors show it consistently beats plain direct workspace interaction and model-internal record-keeping baselines across three benchmarks and three LLMs.

arxiv.orgpaperAdded 8 Oct 2026, published 7 Oct 2026🧠 AI

This paper introduces a detector for adversarial images targeting frozen vision-language models (VLMs) by building a 'response profile' from how an image interacts with a bank of general semantic prompts — using category-level statistics, inter-prompt relationships, deviation from clean reference distributions, and stability under weak image transforms, then classifying that compact profile with a lightweight model. Evaluated across multiple public datasets, several CLIP-style visual backbones, and gradient-based, optimization-based, automated, and spatial attacks, the method outperforms embedding-geometry baselines, generalizes partially to unseen attacks, and is tested against detector-aware adaptive attacks.

arxiv.orgpaperAdded 9 Oct 2026, published 7 Oct 2026🛡️ Security

TaoD2C-Bench is a benchmark for evaluating multimodal large language models (MLLMs) on industrial design-to-code (D2C) tasks — converting UI designs into code that satisfies implementation constraints, not just visual fidelity. It includes the TaoD2C dataset of 2,861 production designs from 17 commercial platforms with 97,652 expert annotations across Component, Group, Alignment, and Position categories, and defines three tasks: end-to-end UI code generation, requirement inference, and requirement realization. Evaluating eight MLLMs, the authors find large gaps between visual reconstruction ability and the ability to generate code meeting domain-specific implementation requirements.

arxiv.orgpaperAdded 8 Oct 2026, published 7 Oct 2026🧠 AI

SLDR is a post-fine-tuning defense that protects aligned LLMs from malicious fine-tuning-as-a-service attacks, which can quietly erode refusal behavior while keeping task performance intact. It trains a LoRA recovery adapter only on the layers with the most extreme (signed) safety sensitivity scores, then uses representation-based dynamic routing to activate that adapter solely for malicious queries at inference time. Evaluated across four model architectures, five downstream tasks, and four harmful-content benchmarks, it drives the average harmful score on Llama3.1/SST2 from 11.54 down to 0.08 while preserving downstream accuracy, and stays near zero even at poisoning ratios up to 0.9.

arxiv.orgpaperAdded 8 Oct 2026, published 7 Oct 2026🧠 AI

This arXiv paper shows that large language models can be trained to tolerate unreliable, faulty digital hardware, and that their error resilience improves rather than degrades as model size increases. Using modified neural scaling laws derived from 40,000 GPU-hours of training runs on simulated faulty hardware, the authors argue that LLMs learn to compute within "good" error-correcting codes whose relative overhead stays finite regardless of scale, suggesting a path toward running AI inference on low-energy, unreliable hardware for substantial energy savings.

arxiv.orgpaperAdded 8 Oct 2026, published 7 Oct 2026🧠 AI

This arXiv paper introduces the "package hallucination attack," where an attacker injects malicious prompts into benign agentic-coding rule files (such as .cursorrules) to induce a coding agent to replace legitimate package dependencies with attacker-controlled ones. The authors propose PackHallu, an evolutionary optimization framework that iteratively rewrites injected prompts using trajectory-level feedback and LLM-guided mutations, and show it achieves high attack success rates with strong transferability across multiple LLMs, benchmarks, and agent frameworks.

arxiv.orgpaperAdded 8 Oct 2026, published 7 Oct 2026🛡️ Security

EC-RAG (Event Chain Retrieval-Augmented Generation) is a training-free framework for long-video question answering that organizes video content into an explicit, temporally ordered "event chain" instead of retrieving isolated frames or snippets. It partitions video into semantically coherent segments, fuses multi-modal signals (speech, text, visual cues) at the event level, and retrieves relevant events from the chain to answer queries, working as a plug-and-play addition to existing large video-language models (LVLMs) without extra training. Evaluated on Video-MME, MLVU, and LongVideoBench, it outperforms frame-level retrieval baselines, showing the value of modeling temporal/event structure for long-video understanding.

arxiv.orgpaperAdded 7 Oct 2026, published 6 Oct 2026🧠 AI

This arXiv paper introduces a pre-registered panel of 248 scenarios across five kinds of pressure, each posed to a model twice (as the acting agent and as a third-person judge) with pressure-removed twins and operator-order positive controls, to measure whether models act against their own stated moral judgment. Testing OLMo-3-7B-Instruct, Llama-3.1-8B-Instruct, Tulu 3, and Qwen2.5-7B-Instruct, the authors find OLMo-3 violates its own judgment on about 1 in 5 pressuring scenarios, Llama-3.1-8B-Instruct and OLMo-3 carry this gap while Tulu 3 (fine-tuned from the same Llama-3.1 weights) and Qwen2.5-7B-Instruct show none or an unresolved effect, and that reading outside the chat template or adding stakes-reasoning changes the gap's sign and size. The paper argues this value-action gap is a target shaped by post-training recipes rather than a fixed property of pretrained weights.

arxiv.orgpaperAdded 7 Oct 2026, published 6 Oct 2026🧠 AI

An IACR ePrint paper presenting K-LMS, an experimental variant of the NIST-standardized LMS hash-based post-quantum signature scheme in which the underlying hash is swapped for Korean primitives: the LSH-256 hash function (reference and vectorized) and a Tandem-DM double-block-length construction built from the CHAM, LEA, and ARIA block ciphers. The authors implement a non-invasive backend-swapping bridge over the reference LMS code and benchmark keygen/sign/verify latency, key/signature sizes, memory, and parameter sweeps, finding that hash substitution doesn't change key/signature sizes but that two measurement pitfalls (hash dispatch overhead, and software vs. hardware-accelerated SHA-256 baselines) can shift performance comparisons by an order of magnitude or 6x respectively. They report LSH-256 K-LMS signing on par with SHA-256 LMS in software, and 3.3-3.9x faster than a prior K-XMSS implementation because LMS needs fewer hash calls per Winternitz step than XMSS.

eprint.iacr.orgpaperAdded 8 Oct 2026, published 6 Oct 2026🛡️ Security

test

arxiv.orgpaperAdded 7 Oct 2026, published 5 Oct 2026🧠 AI

MemPilot is a framework that lets LLM agents curate memory on demand rather than query-agnostically, using a multi-step policy trained via reinforcement learning to choose between retrieving from pre-built memory or delegating query-specific curation of raw multimodal history to heterogeneous LLMs and VLMs. The policy controls evidence amount, curation instructions, model choice, and visual access, and is optimized with objective-wise advantage decoupling and prefix-based marginal utility estimation to balance performance, cost, and latency. Tested across five multimodal agent-memory benchmarks, it achieves broader performance-cost-latency trade-off frontiers than existing trade-off-aware baselines.

arxiv.orgpaperAdded 6 Oct 2026, published 5 Oct 2026🧠 AI

FREA is a benchmark of 751 chemical reactions labeled by expert chemists for feasibility verification, drawn from retrosynthesis model proposals, zero-yield experimental records, LLM-generated edits, and five negative-candidate generation methods. The authors show no single verifier type (LLMs prompted with the feasibility criterion, dedicated forward models, or retrosynthesis scorers) dominates across all sources, and that adding generated negative examples to training improves mean AUROC but fails to transfer to retrosynthesis proposal screening. They also release a corpus of over 14 million recorded and generated reactions to support further training and evaluation of feasibility verifiers.

arxiv.orgpaperAdded 6 Oct 2026, published 5 Oct 2026🧠 AI

Lens3D is a method for improving fine-grained object understanding in 3D large language models by using 3D localization to select informative 2D views that are fed to an external vision-language model (the LensUnd pipeline), then distilling that knowledge back into native 3D LLMs via detailed caption supervision (LensDistill), avoiding the need for external VLM calls at inference time. The authors also release LensBench, a held-out evaluation set of 2,068 objects with three reference descriptions each, and show gains in fine-grained object captioning on Video-3D LLM and 3DRS while preserving grounding and scene-level QA performance.

arxiv.orgpaperAdded 6 Oct 2026, published 5 Oct 2026🧠 AI

SpatialChain is a benchmark of 28,350 training and 899 test examples pairing spatially-oriented GQA questions with scene-graph-grounded reasoning chains, designed to audit whether VLMs' correct spatial answers reflect faithful reasoning or linguistic shortcuts. Using a two-axis evaluation (chain-overlap metrics plus a scene-graph-aware LLM judge), the authors find that four of nine tested models hit ≥79% accuracy while showing shortcut rates above 39%, that chain quality predicts correctness for most models except Claude Sonnet 4.6 and InternVL3.5-8B (which fail differently, via terse vs. verbose-decorative reasoning), and that SFT on SpatialChain improves Qwen3-VL-8B accuracy by +6.2pp while cutting its shortcut rate to 22%. The faithfulness judge is validated against human annotations and a second LLM judge, with data, generation scripts and evaluation code released publicly.

arxiv.orgpaperAdded 6 Oct 2026, published 5 Oct 2026🧠 AI

ImproveAnyTask is an autonomous post-training harness that automates adapting general-purpose LLMs to specific tasks under a limited compute budget, structuring the process into error attribution, update-direction selection, and executable model updates via generated training data and configs with small-scale validation before full runs. Across 11 tasks it reports mean gains of 18.29 and 11.97 percentage points on Base and Instruct models respectively, with validated strategies retained for reuse in future adaptation cycles.

arxiv.orgpaperAdded 6 Oct 2026, published 5 Oct 2026🧠 AI

This paper proposes AI-decision checkpoints, explicit moments in business process development where process developers must evaluate whether to adopt, constrain, or reject AI-candidate sub-processes built with LLMs, RAG, or AI agents, based on expected effects on time, cost, quality, flexibility, and legal or organisational constraints. The checkpoints are embedded in a six-module BPM teaching lifecycle covering process modeling, simulation, workflow execution with AI agents, and process mining, piloted via a fictitious customer-onboarding teaching case with formative observations drawn from instructor logs and student artifacts. It addresses a curriculum gap where AI is still treated as a bolt-on rather than a first-class design element in BPM education.

arxiv.orgpaperAdded 6 Oct 2026, published 5 Oct 2026🧠 AI

This paper studies the sharpening hypothesis of RL post-training on LLMs: that it mainly sharpens existing base-model behaviors, boosting pass@1 at the cost of solution coverage (pass@K), and shows this cost extends to agentic, multi-turn tool-use tasks, not just math/coding. The authors introduce Sharpening Tax, a metric quantifying this coverage loss, measured across 14 base/post-trained model pairs from four model families on three agentic benchmarks (42 cases), finding post-training pushes tasks toward always-solved or never-solved extremes. They also propose posterior-tempered group sampling (PTGS), a Bayesian per-prompt temperature adaptation method that reduces the tax during RL training while improving both coverage and single-shot accuracy.

arxiv.orgpaperAdded 4 Oct 2026🧠 AI

RubricArmor is an adversarial framework for generating evaluation rubrics used in rubric-based reinforcement learning of LLMs, designed to preempt reward hacking before RL training begins. It alternates an attack step, which synthesizes adversarial responses that satisfy the current rubric while failing the actual task, with a repair step that revises the rubric to close the exposed gaps while keeping valid criteria intact. The paper reports that rubrics produced this way outperform existing LLM-based rubric generation baselines and lead to more effective downstream rubric-based RL.

arxiv.orgpaperAdded 6 Oct 2026, published 4 Oct 2026🧠 AI

This paper introduces an expert-validated benchmark of 1,042 items spanning ten areas of Colombian law and three question formats (closed multiple-choice, semi-open, open-ended IRAC), used to evaluate 15 proprietary and open-weight LLMs. Closed-question accuracy ranges from 0.577 up to 0.905 (Gemini 3.1 Pro), but free-text factual correctness never exceeds 0.45 for any model, and only about half of the legal norms models cite in free-text answers are correct — the rest are wrong or fabricated. The study also finds a negative correlation (rho = -0.46) between answer relevancy and factual correctness, meaning models sound confident and responsive while frequently being wrong, and argues current LLMs need expert supervision plus source-grounding for reliable legal use.

arxiv.orgpaperAdded 5 Oct 2026, published 2 Oct 2026🧠 AI

ReFract is a benchmark of 150 expert-validated entries testing whether LLM agents can infer a user's role and act only through tools/actions legitimate for that role, a capability the authors term "Perspective Awareness." Entries are grounded in anonymized industrial-maintenance support queries and use constructed Text World Models to simulate the agent's operating environment and generate perspective-aware action trajectories. State-of-the-art LLMs solve at most 69% of tasks, with over half of trajectories containing attempts at perspective-violating actions, exposing this as a largely unsolved axis of agent evaluation relevant to high-stakes, physically-enacted agent deployments.

arxiv.orgpaperAdded 5 Oct 2026, published 2 Oct 2026🧠 AI

This paper introduces ZFO (Zero-and-First-Order optimization), a lightweight fine-tuning framework that decouples search direction from step-size selection: a trusted first-order optimizer picks the direction, while two extra zeroth-order function evaluations along that single dimension build a local curvature-aware model to choose the step size, avoiding a full line search. The authors give theoretical guarantees on finite-difference curvature estimates, near-optimal step selection, and convergence to a stationary-point neighborhood, and show empirically that ZFO improves optimization and final performance over fixed-step first-order baselines across several language models and datasets. Code is released publicly, making it a usable technique for practitioners fine-tuning LLMs who want cheaper adaptive step-size control.

arxiv.orgpaperAdded 2 Oct 2026, published 1 Oct 2026🧠 AI

Where-OPD introduces an on-policy self-distillation method for multimodal LLMs where a teacher model receives textual spatial guidance (object identities and coordinates from procedurally generated synthetic scenes) to locate relevant image regions, while a student learns to reproduce this behavior from just the image and question. The approach is annotation-free and scalable, improving performance on counting, document and chart understanding, and transferring from synthetic training to real-world perception benchmarks (CVBench, V*, ZoomBench, BLINK, HR-Bench, MME-RealWorld) with a 3.23-point average gain.

arxiv.orgpaperAdded 2 Oct 2026, published 1 Oct 2026🧠 AI

LLM2Jev is an architecture-preserving framework that extracts calibrated categorical decisions directly from an LLM's next-token probabilities over bracketed numeric identifiers, rather than generating free-form text. The paper shows Qwen3.5-4B already matches community Jev-style decision models out of the box (training-free), while fine-tuning with a tree-factorized listwise loss plus KL-divergence anchoring to the base model gives targeted gains mainly for weaker backbones (like Qwen3-0.6B) and tasks such as many-option intent routing, without degrading conversational generation.

arxiv.orgpaperAdded 2 Oct 2026, published 1 Oct 2026🧠 AI

This arXiv paper introduces a benchmark for evaluating whether LLM-generated financial analyses of earnings call transcripts are properly grounded in citable evidence. It proposes a numeric evidence evaluation method that assesses groundedness without expert annotation, an automated dataset construction pipeline, and the resulting ECTs-100 dataset built from the top 100 S&P 500 constituents. The study also examines "conscious incompetence" — whether models correctly abstain when evidence is insufficient rather than hallucinating — finding LLMs ground claims well but still struggle with factual correctness and recognizing insufficient evidence.

arxiv.orgpaperAdded 2 Oct 2026, published 1 Oct 2026🧠 AI

SAGE is a statistical acceptance gate for self-evolving LLM agents that replaces the naive accept-if-aggregate-score-improves rule used when an optimizer edits an agents persistent skill document. It uses per-item paired comparison between the current and edited skill on identical validation items, exposing regressions an aggregate score hides, plus a one-sided paired significance test that only commits edits whose wins reliably beat their losses, addressing both permanent regression risk and the optimizers curse from noisy finite validation sets. Across five benchmarks and four backbone LLMs, SAGE cuts regression rates in 19 of 20 settings, for example 36.5 percent to 0 percent on LiveMath and 42.8 percent to 0 percent on OfficeQA with DeepSeek V4, while also achieving the best final score in all 20 settings.

arxiv.orgpaperAdded 30 Sep 2026🧠 AI

This paper studies whether weight tying between input and output embeddings, common in decoder-only LLMs for parameter efficiency, remains beneficial when models are fine-tuned under Differentially Private Stochastic Gradient Descent (DP-SGD). Using GPT2 and DistilGPT2 fine-tuned on SST-2, QNLI, and QQP, the authors find untied embeddings consistently outperform tied ones by up to 4.74 accuracy points, and that untying also enables memory-efficient ghost clipping (since tied embeddings create shared-parameter interactions that break standard ghost norm computation), yielding over 60% lower memory usage. The results suggest standard LLM architectural choices should be revisited for privacy-preserving training settings.

arxiv.orgpaperAdded 1 Oct 2026, published 30 Sep 2026🧠 AI

TraceDance is an automated system that mines real-world agent deployment traces to build targeted benchmarks for user-specified undesirable agent behaviors, using an Anchor-and-Confirm retrieval-plus-LLM-confirmation pipeline and an Anchor Synthesis Loop to generate behavior specifications. Benchmarks evaluate an LLM's next-turn response at a recorded decision point against a behavior-specific rubric, avoiding the need for reference answers or environment replay. Applied to 252,557 sessions in coding and tool-use domains, it produced 107 benchmarks with 4,125 instances, and testing nine frontier LLMs found a mean pass rate of only 26.7 percent, exposing persistent agentic behavior weaknesses.

arxiv.orgpaperAdded 29 Sep 2026🧠 AI

Imagine3D-LLM appends learnable summary tokens after image tokens in an MLLM, decoding them into a compact 3D Gaussian Splatting representation supervised by a photometric reconstruction loss, trained jointly with next-token prediction. This mimics human coarse spatial reasoning (identifying common objects across views, inferring relative geometry) rather than relying on fine-grained pixel correspondence or fused 3D geometry foundation model features, and the paper shows this reconstruction objective also improves cross-frame correspondence in the LLM's image features. The approach outperforms prior methods across multiple spatial reasoning and 3D understanding benchmarks.

arxiv.orgpaperAdded 30 Sep 2026, published 29 Sep 2026🧠 AI

This paper proves that the Wasserstein distance between Laplacian spectra of original and LLM-reconstructed graphs is bracketed by two edge-count quantities, the net change in edge number and the symmetric difference, each scaled by 2/n, showing when these bounds coincide (pure addition or pure deletion of edges) and when they diverge (mixed editing, i.e. both inventing and losing edges). Using this theory, the authors analyze 135 graph reconstructions produced by three open-weight LLMs over 45 synthetic graphs, finding 77 one-sided outputs and 29 mixed outputs with a positive residual, revealing that aggregate distortion metrics mask qualitatively different editing policies across models, from copying the input to hallucination-heavy completion.

arxiv.orgpaperAdded 30 Sep 2026, published 29 Sep 2026🧠 AI

This paper shows that internal expert-routing probabilities in Mixture-of-Experts vision-language models (Qwen3-VL-30B-A3B-Instruct and Gemma-4-26B-A4B-it) carry a strong, low-cost signal for detecting when a model is about to answer a question about an object that isn't actually in the image (a 'target-absence grounding failure'). Linear detectors trained on target-token routing probabilities achieve ROC-AUCs of 0.9988/0.9956 on GQA-Inpaint (retaining 0.78-0.81 on the external OBER dataset), and gating a target-aware review prompt on these detections improves end-to-end accuracy by up to +22% without modifying model weights. The work matters for practitioners building on MoE VLMs because it demonstrates a weight-free, pre-generation hallucination detector that reuses computation the model already produces.

arxiv.orgpaperAdded 30 Sep 2026, published 29 Sep 2026🧠 AI

This paper introduces Divergent Token Confidence (DTC), a method for uncertainty quantification in LLM chain-of-thought reasoning that counts tokens where two models strongly disagree during decoding, measured via Jensen-Shannon divergence between next-token distributions on the same reasoning trajectory. Unlike standard probability-based confidence scores, DTC requires no training, works in both white-box and black-box settings, and its divergent-token count is shown to be strongly negatively correlated with answer accuracy. Across six mathematical benchmarks and multiple model families (including DeepSeek-V3.2), DTC cuts expected calibration error roughly in half compared to full-sequence probability and verbalized-confidence baselines (e.g., 13.0% vs 32.7%-42.4% white-box).

arxiv.orgpaperAdded 30 Sep 2026, published 29 Sep 2026🧠 AI

Telescopic Language Models (TLM) is a training method that produces a single nested-capacity Transformer valid as a language model at every layer-prefix depth, using stochastic prefix supervision (one randomly truncated prefix trained against the full next-token target, plus one full-capacity pass, per step) with no architectural changes. On a 200M-parameter proxy suite trained on 20B FineWeb-Edu tokens, a single TLM run beats fixed-exit "Matryoshka Language Model Suites" (MLMS) by 43-44% on area-under-the-quality-budget-curve while matching them at full capacity, at ~12% lower GPU cost per run, showing the elasticity comes from the training objective rather than the nesting architecture itself.

arxiv.orgpaperAdded 29 Sep 2026, published 28 Sep 2026🧠 AI

Peppy is an AI-assisted workflow that uses LLMs with domain-specific knowledge of optimization theory to discover tight, analytic convergence proofs for first-order optimization algorithms, verifying results through SymPy rather than a full proof assistant like Lean 4. The authors demonstrate the approach by closing several open problems in convergence analysis, including conjectures about Nesterov's Fast Gradient Method (FGM), positioning the workflow as a more structured, domain-specialized alternative to generic LLM-for-math-research pipelines.

arxiv.orgpaperAdded 29 Sep 2026, published 28 Sep 2026🧠 AI

This paper introduces NstAgent (Narrative State Tracking Agent), a training-free agentic framework that lets LLMs maintain a structured narrative state — characters, past events, and future requirements — to sustain consistency when generating full-length novels. The authors extend an existing narrative-consistency benchmark and pair it with a writing-quality benchmark to evaluate stories from 10K to 100K words, showing that NstAgent's consistency and quality do not degrade as length increases, unlike prior methods limited to ~10K-word stories.

arxiv.orgpaperAdded 29 Sep 2026, published 28 Sep 2026🧠 AI