129 results for “Qwen3”

nesformer is a project by Drew Breunig that bakes a working Nintendo Entertainment System 6502 CPU emulator into the weights of Qwen3.5-4B, so NES ROMs such as Super Mario Bros run (very slowly) inside the model. It follows Divyansh Agrawal's qwenputer. Simon Willison relays the Opus 5.5 prompts Breunig used, and notes that Opus found Christopher Pow's CPU conformance tests.

simonwillison.netarticlePublished and added 30 Oct 2026🧠 AI

Thread by Elvis Saravia (omarsar0) summarising an Anthropic-led paper on pretraining data poisoning of CLI agents. Poisoning 0.1% of pretraining data planted a remote code execution backdoor in Qwen3 models (0.6B, 1.7B, 4B) trained from scratch. It survived SFT, DPO and RL, and fired on about 92% of triggered prompts at 4B. At a 0.0005% poison rate (about 2,500 documents) it still fired on 1.4% of prompts.

x.comthreadAdded 31 Oct 2026, published 30 Oct 2026🛡️ Security

DecoyCount is a benchmark from Elorian and Mercor with 1,660 items that tests whether models count objects or fall back on visual shortcuts. Frontier models scored below 50%: GPT-6 Astra led with 43.17%, ahead of Gemini-3.8-flash (32.71%), Qwen3.8-max (28.31%) and Claude-Fable-5.1 (25.22%). CEO Andrew Dai says Elorian's Glimpse model scored 50.48% in a single internal run without tools, and contrasts visual reasoning with world models such as Atlas and Genie 4.

latent.spacearticleAdded 28 Oct 2026, published 29 Oct 2026🧠 AI

Turing LLM post-training combines calibrated multi-teacher on-policy distillation (C-MOPD) with an execution path for dynamically routed mixture-of-experts models. C-MOPD matches each teacher on a local Top-K support with teacher-wise loss normalization and clipping. On Turing-20B-A2B it scores 0.9988 on four-domain integration, beats Qwen3-8B on math and code (85.92 vs 80.33 mean), and cuts consolidation step time from 398.7 to 139.3 seconds in a PPU profile.

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

REALM (Required Execution Assets Linked to a Mission) is an approach to applying least privilege to LLM agents by identifying the assets a task needs, including ones discovered at runtime. It models assets as Dependency Graph nodes, separates discovering an asset from authorizing it, and exposes resource-bound Partial APIs inside a Task Container. On AgentDyn with qwen3.5-flash, attack success rate falls from 16.6% to 0.2%, with 40.0% task completion on clean tasks and 36.4% on injected cases.

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

Qwenputer is a stock Qwen3.5 4B with a 32-bit RISC-V CPU (RV32IM plus Zba and a bignum unit) compiled directly into added layers, with no training. Gated DeltaNet states act as exact RAM via gate 1, beta 1 and one-hot keys. It scores 97.5 to 100% on DeepMind arithmetic sets, and 100% on NUPA long-integer arithmetic. It also has an experimental in-model filesystem. The checkpoint is 18.8 GB and needs the RvChat driver.

divagr.comwriteupPublished and added 29 Oct 2026🧠 AI

AnySpeechRecognizer and AnyDecisionModel are two Swift packages from Hugging Face. The first mirrors Apple's SpeechAnalyzer API with swappable engines (NVIDIA Parakeet, Alibaba Qwen3-ASR on Core ML) and adds Sortformer speaker diarization. The second returns typed yes/no, choice and score answers with probabilities via MLX, Core AI, llama.cpp, Jev or Ollama. The post gives M4 Max timings, model sizes and calibration options.

huggingface.coarticlePublished and added 29 Oct 2026🧠 AI

AINews issue of 28 October 2026 from Latent Space. It leads with Felix Rieseberg's post on why Anthropic dropped coding interviews ("Hunger beats tenure"). It then recaps Genie 4, Axiom's Lean formalization of CFSG, claimed ETH refutations by OpenAI and Anthropic, Artificial Analysis Index v5, FrontierCode 2.0, Liquid LFM3, the Qwen3.8 phone-home dispute, and Anthropic's S-1. Reddit recaps are included; the page was truncated before the end.

latent.spacearticlePublished and added 28 Oct 2026🧠 AI

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.

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

EnvSched is an online scheduler for agentic reinforcement learning that picks which environment-task pairs to roll out next. It builds a dual-relation graph over environments and tasks, propagates past rollout statistics to predict within-task reward variance, and adds environment quotas and a similarity penalty. On Qwen3-8B it scores 48.81 average on BFCL-v3, tau-Bench and ACEBench-Agent after 40 steps, above random selection at 120 steps.

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

LFM3 is Liquid AI's open-weight family of six ~2B-parameter models (text, vision-language, audio, omni and two d1 decision models), all branched from LFM3-1.7B-Base. The base model uses a tri-hybrid backbone of 13 gated short convolutions, 5 sliding-window attention layers and 6 global attention layers, pre-trained on 22T tokens with a 256K context. The post gives transformers snippets, benchmark comparisons against LFM2.5 and Qwen3.5, and a robotics fine-tuning example.

huggingface.coarticlePublished and added 27 Oct 2026🧠 AI

This paper studies a cost-aware, two-action RAG routing policy that chooses between dense retrieval plus reranking and a hybrid lexical-dense retriever, using a regularized linear model to predict per-question evidence gain. On 60 held-out QASPER articles (188 questions), it triggers hybrid retrieval on 24.5% of questions and matches always-hybrid Hit@10 (78.70%). Answers from 4-bit Qwen3-4B show no clear correctness or faithfulness gain, and the assessments are model-assisted.

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

Walkthrough of running OpenCode against Qwen3.5 9B served by Ollama on a Mac, with Wi-Fi switched off. The author fixes a one-line accumulation bug in an expense calculator and confirms all five unittest tests pass. It covers the opencode.json setup, memory budgeting, permission prompts and a pytest-versus-unittest detour. Two further exercises (CSV report, log analyzer) were not yet run offline.

latent.spacewriteupPublished and added 27 Oct 2026🧠 AI

DCKV is a differentiable codec for the prompt KV cache of hybrid language models, which keep a KV cache only in their few full-attention layers. Transform, quantizer and entropy models (conditioned on prompt tokens and already decoded neighbouring positions) are trained end to end against a frozen model. On Qwen3.5-2B it stores a token in over 55 times fewer bytes than bf16 with at most 1.8 points lost on three long-context benchmarks, and over 100 times fewer with at most 3.6 points lost.

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

Nanbeige4.3-33B is a 33.1B dense, Llama-style open model under Apache-2.0, released with a base model on Hugging Face. In Nanbeige's own table it beats Qwen3.8-27B on Terminal-Bench 2.1 (80.9 vs 73.0), DeepSWE (65.5 vs 42.5) and Automation Bench, but trails on SWE-Bench Pro, GDPval and GPQA Diamond. BF16 weights are about 66 GB, split across 4 GPUs in the vLLM and SGLang examples. No independent evals exist yet.

x.comthreadAdded 26 Oct 2026, published 25 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

An X post showing how to have Codex train a small model on Apple silicon through the mlx-lm-lora MCP server and skill. It gives five setup steps (pip install with the mcp extra, npx skills add, locate mlx-lm-lora-mcp, codex mcp add). The prompt runs four pre-training trials on tiny-random/qwen3 with the mlx-community/recycling_the_web-1k dataset. Codex picks the optimizer and learning rate itself, and the result is a table with validation loss and the best adapter path.

x.comthreadAdded 25 Oct 2026, published 24 Oct 2026🧠 AI

Building Reasoning Models from the Ground Up is a series of eight coding videos (about 11 hours) by Sebastian Raschka, accompanying his book Build a Reasoning Model (From Scratch). Starting from a pre-trained Qwen3 0.6B model, it implements KV-cached generation, a MATH-500 verifier, self-consistency and self-refinement, simplified GRPO for RLVR, and distillation from DeepSeek R1 answers in Python and PyTorch.

magazine.sebastianraschka.comcoursePublished and added 24 Oct 2026🧠 AI

SemiAnalysis reviews AWS Trainium3 and the new TorchNeuron native PyTorch backend after a few weeks of hands-on use. It reports TorchTitan Qwen3-32B results on Trainium2 (11,099 tokens/s eager at 23.89% MFU, 14,156 compiled at 30.47%), a 27-minute gpt-oss compile time, 89 of 96 FlexAttention cases passing, matmul comparisons with GB300 and B200, and a positive view of the Neuron Explorer profiler.

newsletter.semianalysis.comarticleAdded 29 Oct 2026, published 23 Oct 2026🧠 AI

ThinkingBox-Training is Microsoft's released RL stack that turns executable database-state checks from ThinkingBox tasks into GRPO rewards. Fifty GRPO steps on 187 curated tasks raised Qwen3.8-27B pass@1 from 51.70% to 60.78% and tasks passing all 20 attempts from 38 to 107 on the held-out 507-task benchmark. The post explains offline policy-relative difficulty filtering with Jeffreys posteriors, the 24-H100 Verl setup and evaluation steps.

huggingface.cowriteupPublished and added 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

RIFT is a family of selection-based KV cache compression strategies for speculative-decoding draft models. It targets "decode staleness", where prefill-focused selection loses relevance as output grows, by balancing prefill information with recent generated tokens, refreshing the selection adaptively, and training drafters with sparse attention masks. Qwen3.5-4B drafting for Qwen3.5-122B keeps over 98% of full-cache acceptance length at 32K input, 4K output and an 8K KV budget.

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

VocabX is a vocabulary adaptation method from Dharma-AI that selects sub-word and multi-word "macro" tokens by serving cost instead of frequency. On a Qwen3-4B Portuguese essay grader, response length fell from 350 to 101 tokens, the LM head shrank from about 152,000 to 65,000 rows, and measured cost dropped 77.0% on an RTX 5090 with vLLM. The bench score rose from 0.703 to 0.810. It beat FVT, MWT and in-place BPE across 15 model and domain combinations.

huggingface.coarticlePublished and added 21 Oct 2026🧠 AI

Hirundo's case study on Qwen3-Coder-30B-A3B-Instruct, tested with the BaxBench SecretStorage task in Flask. Apps for a US defense contractor were 19% more likely to be hackable than for a Chinese one, and Hong Kong democracy group apps leaked secrets 28% more often. A security instruction did not close the gap, but Hirundo's weight-edited "Westernized" model did. Semgrep with PurpleLlama rules flagged 2 of 1,840 apps; BaxBench attacks broke 1,144.

hirundo.ioarticleAdded 22 Oct 2026, published 21 Oct 2026🛡️ Security

Hirundo case study on Qwen3-Coder-30B-A3B-Instruct using the BaxBench SecretStorage task. Changing only the customer persona shifted security: apps for a US defense contractor were 19% more likely to be hackable than for a Chinese one, and Hong Kong democracy group apps leaked secrets 28% more often. Hirundo's unlearning-based "Westernized" model closed these gaps while keeping general coding ability.

hirundo.iowriteupAdded 22 Oct 2026, published 21 Oct 2026🧠 AI

Hirundo tested Alibaba's Qwen3-Coder-30B-A3B-Instruct by having it build 1,840 web apps for a US defense contractor, a Chinese defense contractor, a Hong Kong advocacy group and a neutral club. BaxBench broke into 1,144 apps; 70% of US contractor apps were hackable versus 59% for the Chinese one, and 74% when told to write secure code. Hirundo says removing the model's CCP alignment from its weights closes the gap.

securityweek.comarticlePublished and added 21 Oct 2026🛡️ Security

Hirundo tested Alibaba's Qwen3-Coder-30B-A3B-Instruct by having it build 1,840 web apps for a US defense contractor, a Chinese defense contractor, a Hong Kong advocacy group and a neutral club, scored with Semgrep, PurpleLlama rules and BaxBench. 70% of apps for the US contractor were hackable versus 59% for the Chinese one, and a security instruction did not close the gap. Hirundo says its unlearning of CCP alignment removed the bias.

securityweek.comarticlePublished and added 21 Oct 2026🧠 AI

Protection Is Not One Event is a survey and human-coded evidence map of 118 RAG defense papers, grouped by measured endpoint into a seven-family taxonomy (kappa=0.867 across 32 annotators). It finds that only 17.8% of studies use content disclosure as their primary endpoint, and that most evaluations use general QA data and non-adaptive attacks. In a testbed on Qwen3-8B, an 8-gram guard cut copying to 0.0%, yet a three-query adaptive attacker still recovered 48.3% of hidden facts. Taxonomy, dataset, coding protocols and testbed are released.

arxiv.orgpaperAdded 24 Oct 2026, published 21 Oct 2026🧠 AI

Survey and human-coded evidence map of 118 RAG defense papers, organized by measured endpoint across a seven-family taxonomy (content disclosure, formal privacy proxies, access control, corpus-use detection, pipeline integrity). Only 17.8% of studies use disclosure or reconstruction as the primary endpoint, and most use non-adaptive attacks on general QA data. A testbed shows a normalized 8-gram guard cuts 20-token copying to 0.0%, yet a three-query adaptive attacker recovers 48.3% of hidden facts on Qwen3-8B.

arxiv.orgpaperAdded 24 Oct 2026, published 21 Oct 2026🛡️ Security

PyTorch 2.15 MPS release post from the Hugging Face PyTorch team, covering the Apple silicon backend. It adds float8_e4m3fn tensors with F.scaled_mm, a native grouped matmul that speeds MoE decoding 5 to 6.5x (Qwen3.6-35B-A3B from 9.4 to 61.3 tokens/s on an M3 Ultra), and an 8x faster softmax. Wan2.2 VAE decode peak memory falls from 270 GiB to 8 GiB. It previews 2.16 attention on Metal.

huggingface.coarticlePublished and added 20 Oct 2026🧠 AI