GPT-Rosalind-6 System Card Addendum is OpenAI's safety documentation for a life-sciences model that shares weights with GPT-6 Astra. It rates the model High in Biological and Chemical and Critical in Cybersecurity under the Preparedness Framework. It reports biology evaluations, trusted-access controls, biosecurity refusal pass rates, a historical cyber jailbreak regression (14.3% task success), and sabotage and monitor-evasion results.
GPT-Rosalind-6 System Card Addendum describes OpenAI's life-sciences reasoning model, which uses the same weights as GPT-6 Astra and is offered only through trusted access. It reports life-science benchmark gains over GPT-Rosalind-5.5 (MedChemBench v7 49.6% vs 28.9%), High-capability Biological and Chemical Preparedness results, and below-threshold Critical bio design evaluations. It also covers access controls, refusal evaluations, cyber jailbreak regression and alignment tables.
GPT-6 AstraGPT-Rosalind-6GPT-Rosalind-5.5
OpenAIPreparedness FrameworkMedChemBench v7+16
ai-safety, evals, modelsΒ· Published and added 30 Oct 2026
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.
vision-language modelQwen3-VLGLM+2
GLM-4.5VGPT-5.4Claude-Opus-4.7+4
evals, multimodal, researchΒ· Added 31 Oct 2026, published 29 Oct 2026
High Precision Noise Cleaning for CKKS is an IACR ePrint preprint (2026/2675) by Zhihao Li, Cheng Hong and Tao Wei. It composes lower-degree Hermite polynomials to reach output noise bounds such as 2^-300 for noise flooding, and uses this to build an asynchronous threshold CKKS scheme with strong simulation security in the random oracle model. In OpenFHE, binary-message cleaning falls from 9.63 s to 1.54 s.
Colin O'Flynn's paper shows that the flash accelerator in most microcontrollers produces the same timing collisions as a cache, so cacheless Cortex-M devices remain open to cache-collision attacks on AES. AES keys were recovered from as few as 750 traces across six cryptographic libraries, eight architectures and over twenty devices. It also attacks SPI bootloaders, Ethernet, CTAP2 over USB and 802.15.4, and releases the open-source tool FlashWhisperer.
FlashWhispererARM Cortex-M4SPI+1
cache-collision timing attackColin O'FlynnCTAP2+2
Hardware & RFΒ· Added 31 Oct 2026, published 29 Oct 2026
Addendum to the GPT-6 Luna system card for the model update of 29 October 2026 (the title says October 28). OpenAI keeps the High rating for cybersecurity and biological/chemical capability under its Preparedness Framework. It reports ExploitBench 46.19% and SEC-Bench Pro 51.75% (GPT-6 Astra: 85.4%), higher multiturn jailbreak defender success than the October 7 version, and GPT-Red prompt-injection results. It also reports weaker factuality and dual-use biology refusal.
AskStream is a multi-turn benchmark for streaming video understanding that adds a third proactive action, ask, alongside wait and answer, for cases where video or user information is incomplete. It has 438 videos, 2,714 manually annotated frame-level time windows (wait, ask, answer), four task categories and eight subtasks. The strongest model reaches only 52.18% on the proposed Ask-F1 metric, in both multiple-choice and open-ended settings.
multimodal large language model
AskStreamAsk-F1arXiv
evals, multimodal, researchΒ· Added 30 Oct 2026, published 28 Oct 2026
This paper measures job-based doubt in LLM fact-checkers: judging a claim false more often when the poster has a lower-status job. It compares base and post-trained checkpoints of six model families on fact-checked claims with randomised poster profiles. Only the open TΓΌlu-3 recipe (OLMo-2 32B, TΓΌlu-3 70B) creates the effect, under one percentage point but reliable. In OLMo-2 32B the flip comes from one residual-stream direction and appears at supervised fine-tuning, confirmed on 400 new claims. None of seven commercial models doubts lower-status posters more.
LLMTransformer
OLMo-2 32BTΓΌlu-3TΓΌlu-3 70B+3
research, ai-safety, trainingΒ· Added 30 Oct 2026, published 28 Oct 2026
This paper measures how much a model's final answer depends on its own written chain of thought, using a synthetic box-swap task with exact ground truth. It edits one step in the model's chain to a wrong but valid state and checks whether the answer follows. Across 25 open-weight models, reliance rises with size and saturates (Qwen3 from 16-18% at 0.6B to 97-100% at 14B). Reliance tracks capability (r = 0.99) and is unchanged by post-training.
Qwen3GemmaLlama+2
Gemma 3Llama 3arXiv
research, ai-safety, evalsΒ· Added 30 Oct 2026, published 28 Oct 2026
This paper treats LLM reasoning effort as a credence good and measures the counterfactual effect of effort levels on proprietary and open-weight models across four benchmarks, including long-horizon agentic tasks. Higher effort often fails on tasks lower effort solved, and on easy tasks the top level costs up to 42x the lowest. Starting at the lowest effort and escalating only after a verified failure matches or beats fixed top effort, usually at lower cost for proprietary models.
LLM
Hugging Face
research, evals, agents, ai-policyΒ· Added 28 Oct 2026
Locating Sycophancy Across Training Stages in Open Language Models measures sycophancy in 13 open checkpoints from 5 lineages, from base model through supervised fine-tuning and preference optimization to release. It finds two behaviors with different origins: deference to a claim in the prompt exists in every base model, while capitulation under user objection is largely absent from base models and present in every post-trained checkpoint. Post-trained models often drop firmly held answers as readily as loosely held ones when the objection names an alternative.
LLM
arXiv
ai-safety, research, evalsΒ· Added 30 Oct 2026, published 27 Oct 2026
OpenGray is a benchmark that adds tasks with known correct decisions to the OpenKBP-Opt head-and-neck dataset, including two overlap contradictions with provably incompatible dose goals and a feasible control. Four planning systems, from a rule-based script to a full LLM agent, were tested on 30 patients with Terra and Ministral 3 8B. Decision accuracy reached 95.4% and 92.1%, yet only 50% and 25% of overlap contradictions were escalated. SAGE logs of 39 radiosurgery runs showed an agent that lacked an action for stopping.
LLM
OpenGrayOpenKBP-OptSAGE+3
evals, agents, researchΒ· Added 28 Oct 2026, published 27 Oct 2026
An empirical study of security concerns in the OpenClaw AI personal assistant framework, using 25K+ posts from 13 subreddits (162 security-relevant) and 358 security-tagged GitHub issues (180 retained). It maps 521 CWE instances across 59 weakness types, concentrated in access control, sensitive-data exposure, injection and unsafe execution. It extracts 543 practitioner mitigations into seven strategies and reconstructs a sensitive-data exposure path with dummy secrets.
OpenClawGitHubLLM
prompt injectionCWEtrust boundaries+1
AI SecurityΒ· Added 28 Oct 2026, published 26 Oct 2026
Scope-Gap Injection (SGI) is an indirect prompt injection method for multi-stage LLM-agent safety pipelines, tested in the IJCAI 2026 OpenClaw Agent Security Attack Competition. The visible task and skill stay benign while target identifiers, bypass parameters and authorization steps sit in an attachment written as a business worklist, which the S2 LLM judge appears not to inspect. SGI bypassed all three stages (static scan, LLM judge, live audit) for 5 of 19 targets, best F = 0.385. Attachment placement had 78% S2 acceptance versus 23% in-skill. The paper closes with defenses that align evaluator scope with agent channels.
Addendum to the GPT-6 Astra system card for OpenAI's GPT-6.1 Terra, rated Critical in cybersecurity and High in biological and chemical capability under the Preparedness Framework. It reports multiturn jailbreak and prompt-injection results, alignment tests (auto-review circumvention, ExploitGym honeypot, monitor evasion), and cyber scores: 97.10% ExploitBench, 24.5% on the internal recent-vulnerability port, 5 of 22 Sandbox Bench targets.
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.
LLMpii-guard
AI SecurityΒ· Added 27 Oct 2026, published 25 Oct 2026
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.
LLMpii-guard
Vladimir BeskorovainyiarXiv
evals, llm-apps, researchΒ· Added 27 Oct 2026, published 25 Oct 2026
Gao and Hu's IACR ePrint 2026/2622 shows that if the matrix multiplication exponent satisfies Ο β€ 9/4 β Ξ·, heuristic lattice sieving runs in (3/2 β Ξ΄)^(n/2+o(n)) time under the spherical heuristic from the BDGL analysis. It builds on OpenAI's recent result that Ο β€ 9/4. With known matrix multiplication algorithms, the method already gives subquadratic search inside BDGL buckets. Only the abstract is available, so concrete parameters are not shown.
IQuest-Q1 is an open-source sparse Mixture-of-Experts model with 320B total and 15B activated parameters, built for agentic coding in CLI systems. Pre-training combines Stable SplitNorm, Muon and muP with a two-stage curriculum that raises the coding-data share in stage two. Mid-training extends context, and post-training mixes SFT, domain-specialized RL and MOPD across scaffolds, followed by model merging.
Mixture of ExpertsStable SplitNormMuon+3
IQuest-Q1Hugging Face
models, training, coding-agents, researchΒ· Added 24 Oct 2026
This paper studies demographic bias in LLM-emulated electorates, using three U.S. presidential elections and comparing one closed-source and one open-source LLM. Removing gender from the voter profile cuts the largest group error by 22-42%, and removing education cuts it by 64%. Removing self-reported race raises mean group error by over 40%. Repeated queries and fairness instructions do not reduce errors and can double them.
LLM
arXiv
research, ai-safety, evalsΒ· Added 27 Oct 2026, published 23 Oct 2026
OWLS (Open-world Language-conditioned Speed Fields) plans directly over a vision-language model's dense feature map. A text prompt describing traversable terrain scores every pixel of one camera image. The scores are projected onto the ground, adapted with two morphological operations, and used as the speed field of a minimum-time planner. It needs no training, terrain labels, map or extra sensors. It was tested on a physical rover, on simulated Mars and Moon rovers over 200 procedural terrains, and on real planetary imagery.
vision-language model
OWLSarXiv
research, multimodalΒ· Added 26 Oct 2026, published 23 Oct 2026
DeepFaune-UK is a fine-tune of the open DeepFaune camera-trap classifier on 28 UK classes, paired with MegaDetector v6. On 20,212 MammalWeb images it reached 88.8% strict accuracy against 56.5% (DeepFaune v1.5.0) and 59.0% (SpeciesNet). Re-applying the recipe to the DINOv3-based classifier gave 90.6%. Fine-tuning on about 150 crops per class matched full-data out-of-distribution accuracy, and a DINOv2 run took about 22 minutes on under 4 GB of GPU memory.
DeepFauneMegaDetectorSpeciesNet+2
DeepFaune-UKMammalWeb
research, training, multimodalΒ· Added 26 Oct 2026, published 23 Oct 2026
MCP-Injector is a protocol-aware fault-injection middleware for Model Context Protocol tool calls, with permanent, transient and intermittent faults on seedable schedules. Its fault model comes from 57,240 issue and PR threads across 1,774 open-source MCP repositories. In a 1,350-cell experiment, faults cut task-completion score by 60% (8.08 to 3.23 of 10), and the 120B gpt-oss model was no more resilient than the 20B. Agent stacks differed in how well they absorbed faults.
Model Context ProtocolOpenAI Agents SDKLangChain+2
MCP-Injectorgpt-oss-120bOpenAI
mcp, evals, agents, researchΒ· Added 26 Oct 2026, published 23 Oct 2026
When Agents Fail: Model or Harness--Who's to Blame? proposes five measures for evaluating agent harnesses together with their underlying models: Valid Completion Rate, Model-Harness Fit, Cross-Model Fit, Interaction Residuals and Token Load. It tests four harnesses (OpenCode, Codex CLI, Pi CLI, DeepAgents) with four models on six benchmarks, across 111 settings and 52,168 trajectories. Harness rankings change by model and benchmark. OpenCode ranks first in cross-model performance.
OpenCodeCodexPi+2
DeepSeek V4arXiv
evals, agents, researchΒ· Added 26 Oct 2026, published 23 Oct 2026
MART-GRPO is a variant of group-relative policy optimization for vision-language-action (VLA) reinforcement learning. It uses a LoRA-fine-tuned VLM to find the last visible error without recovery in a failed rollout video, keeps failure feedback up to that point, and drops it afterward. Martingale stopping conditions show this preserves the expected GRPO gradient and can reduce covariance. With OpenVLA-OFT on RoboTwin 2.0, it raises success rates over full-trajectory GRPO across multiple tasks.
GRPOLoRAvision-language-action model+2
RoboTwin 2.0arXiv
research, training, multimodalΒ· Added 26 Oct 2026, published 23 Oct 2026
Progress Probes (P^2) trains linear probes on a vision-language-action model's own activations, using demonstrations, to estimate task progress, then flags executions that deviate from expected progress. Tested on the open-source pi0.5 across simulated and real-world manipulation tasks, it matches or beats VLM-based monitors and failure-data-trained methods at failure detection, without an external foundation model.
CellTrack-Bench is a benchmark that tests vision-language models on microscopic cell tracking and compares them with specialised trackers. It has 1,537 problems covering 9,041 cell-linking decisions, drawn from 15 cell-tracking datasets across four imaging modalities, with a hard subset built from failures shared by distinct algorithmic trackers. The authors compare environment harnesses, benchmark 9 closed-source and 13 open-weight models, and analyse failure modes.
vision-language modelLLM
CellTrack-BencharXiv
evals, multimodal, researchΒ· Added 24 Oct 2026, published 22 Oct 2026
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.
Qwen3LlamaLLM
ToolBenchClaudeChatGPT+1
research, training, agentsΒ· Added 24 Oct 2026, published 22 Oct 2026
Linear cryptanalysis of FRAST, a TFHE-friendly stream cipher, in the average key setting. It fixes the false-positive problem left open in the earlier attack by Bak et al., showing key differences fall into five indistinguishable classes per round. The Contract algorithm recovers the master key for 10 rounds and, on weak keys of density 2^-11.8 to 2^-5.4, reaches full FRAST within the claimed security.
FRASTTFHE
Contract algorithmBak et al.linear cryptanalysis+1
Paper (arXiv, 21 October 2026) on structural false negatives in LLM vulnerability discovery harnesses. Starting from Knostic's OpenAnt and raptor, then 10 more open-source harnesses, the authors map 13 points where code selection, filtering, and finding storage or merging drop vulnerabilities. One file was dropped by six harnesses before the model saw it, five without notice. It covers attacker evasion, content-steered verdicts, prompt injection, and design principles and tests.
OpenAntraptorLLM
KnosticPtacek and Newshamprompt injection+1
AI Security, AppSecΒ· Added 24 Oct 2026, published 21 Oct 2026
deploymentsafety.openai.compaperπ‘οΈ Security
