https://intelsi.si/ Discover frontier AI models, autonomous agents, research assistants and AI safety tools. A curated, searchable directory with a live research feed. en Tue, 06 Oct 2026 00:40:41 +0000 https://www.theguardian.com/technology/2026/oct/03/openai-safety-leader-quits-warning-ai-companys-culture-is-broken [Hacker News] 268 points · 3 comments Hacker News Sat, 03 Oct 2026 22:18:13 +0000 intelsi.si:hn-49948332 https://www.science.org/content/article/exclusive-ai-agent-emailed-hundreds-researchers-help-it-told-us-why [Hacker News] 50 points · 80 comments Hacker News Sat, 03 Oct 2026 10:07:08 +0000 intelsi.si:hn-49942865 https://royapakzad.substack.com/p/multilingual-ai-agents [Hacker News] 44 points · 1 comments Hacker News Fri, 02 Oct 2026 20:39:05 +0000 intelsi.si:hn-49938326 https://arxiv.org/abs/2610.03715v1 [arXiv] We introduce 4DCodeBench, a benchmark for 4D inverse graphics through code generation, in which agents reconstruct dynamic scenes from video as executable graphics programs. To accomplish this, agents must translate visual observations into compact representations of scene struct… arXiv Fri, 02 Oct 2026 17:58:49 +0000 intelsi.si:arxiv-2610.03715v1 https://arxiv.org/abs/2610.03634v1 [arXiv] Terminal-using agents benefit from reinforcement learning (RL) in coding, debugging, and other multi-step terminal tasks. In these tasks, later commands often depend on information or intermediate results produced by earlier commands. However, existing trajectory-level and step-l… arXiv Fri, 02 Oct 2026 17:24:48 +0000 intelsi.si:arxiv-2610.03634v1 https://arxiv.org/abs/2610.03631v1 [arXiv] Designing a scientific instrument tests whether language-model agents can do physics rather than recall it, provided the grading cannot be argued with. We introduce NeutronGym, to our knowledge the first executable environment for neutron instrument design: agents build instrumen… arXiv Fri, 02 Oct 2026 17:23:52 +0000 intelsi.si:arxiv-2610.03631v1 https://arxiv.org/abs/2610.03591v1 [arXiv] Understanding and assessing natural hazards is essential for disaster preparedness and risk reduction. Recent advances in large language models have spurred growing interest in AI agents for hazard analysis, particularly their ability to integrate scientific data, models, and too… arXiv Fri, 02 Oct 2026 16:59:19 +0000 intelsi.si:arxiv-2610.03591v1 https://arxiv.org/abs/2610.03585v1 [arXiv] Security benchmarks for LLM-based agents often report the attack success rate (ASR) as a measure of model robustness and use these scores to compare different models and defense mechanisms, assuming that they describe the security of the agent. In this paper, we explore whether i… arXiv Fri, 02 Oct 2026 16:55:50 +0000 intelsi.si:arxiv-2610.03585v1 https://arxiv.org/abs/2610.03574v1 [arXiv] We introduce HyperBrowseComp, a multilingual and multimodal browsing benchmark comprising 423 manually authored and human-validated questions across 13 languages, written by native or highly proficient speakers. Questions are designed to be extremely challenging. Each question ta… arXiv Fri, 02 Oct 2026 16:48:49 +0000 intelsi.si:arxiv-2610.03574v1 https://arxiv.org/abs/2610.03564v1 [arXiv] Financial AI agents must do more than retrieve facts: investment workflows require correct quantitative execution, reliable use of procedural resources, and auditable structured outputs. We introduce FinSkillBench, an evaluation suite of 2,603 point in time episodes across 12 sub… arXiv Fri, 02 Oct 2026 16:44:04 +0000 intelsi.si:arxiv-2610.03564v1 https://arxiv.org/abs/2610.03548v1 [arXiv] Generating progressively harder reasoning problems requires synthesis procedures that adapt as the task distribution evolves. Existing task-level recursion reuses generated problems as seeds but leaves the construction harness unchanged. We present task-harness co-evolution, a fr… arXiv Fri, 02 Oct 2026 16:30:15 +0000 intelsi.si:arxiv-2610.03548v1 https://arxiv.org/abs/2610.03519v1 [arXiv] A reasoning model asked whether a causal effect is recoverable from observational data can fail in two ways: it refuses an identifiable query or answers a nonidentifiable one. The latter is more consequential, as no observational data can validate the claimed formula. Measuring t… arXiv Fri, 02 Oct 2026 16:12:21 +0000 intelsi.si:arxiv-2610.03519v1 https://arxiv.org/abs/2610.03515v1 [arXiv] On-policy self-distillation (OPSD) uses reference solutions as privileged hindsight to supervise student-generated reasoning trajectories. However, reference-based guidance may explain a correct solution without addressing why the student's own reasoning fails. This reasoning mis… arXiv Fri, 02 Oct 2026 16:07:39 +0000 intelsi.si:arxiv-2610.03515v1 https://arxiv.org/abs/2610.03509v1 [arXiv] Chain-of-thought (CoT) reasoning allows humans to inspect how large language models reach their answers, and oversee model behaviour. This reasoning comes at an increased inference cost, motivating efficient methods that train models to solve tasks using fewer tokens. However, a … arXiv Fri, 02 Oct 2026 16:04:08 +0000 intelsi.si:arxiv-2610.03509v1 https://arxiv.org/abs/2610.03448v1 [arXiv] LLM agents increasingly screen tool outputs with small prompt-injection detectors, and teams choose among detectors by their scores on public benchmarks. We ask whether those scores predict how a detector behaves inside an agent. We replay the ground-truth tool calls of two agent… arXiv Fri, 02 Oct 2026 15:30:11 +0000 intelsi.si:arxiv-2610.03448v1 https://arxiv.org/abs/2610.03356v1 [arXiv] Large Language Model (LLM) agents are increasingly deployed in high-stakes settings such as industrial maintenance and equipment fault troubleshooting, where workers occupy a variety of roles. A capable agent must therefore act in a way that is calibrated to user's role: taking a… arXiv Fri, 02 Oct 2026 14:21:26 +0000 intelsi.si:arxiv-2610.03356v1 https://arxiv.org/abs/2610.03329v1 [arXiv] Large language models are increasingly used where small syntactic errors matter, yet character-level reasoning is still evaluated mostly through isolated probes and aggregate accuracy. We introduce SyntaxBench, a diagnostic benchmark and statistical evaluation framework for chara… arXiv Fri, 02 Oct 2026 14:02:56 +0000 intelsi.si:arxiv-2610.03329v1 https://arxiv.org/abs/2610.03326v1 [arXiv] Diffusion language model (dLLM) compression faces a known challenge because calibration is typically performed on clean, fully visible activations, whereas inference traverses partially masked intermediate states. For low-rank compression, this raises two questions. First, can lo… arXiv Fri, 02 Oct 2026 14:00:55 +0000 intelsi.si:arxiv-2610.03326v1 https://arxiv.org/abs/2610.03315v1 [arXiv] Trajectory evaluation is essential for improving the reliability of LLM-based agents, but production use makes it expensive to run repeatedly. Modern agents generate long traces containing tool calls, observations, retries, and external outputs, while not all raw tokens are equal… arXiv Fri, 02 Oct 2026 13:51:06 +0000 intelsi.si:arxiv-2610.03315v1 https://aweb.ai [Hacker News] 42 points · 28 comments Hacker News Thu, 01 Oct 2026 22:02:31 +0000 intelsi.si:hn-49927587 https://erictopol.substack.com/p/loss-of-cell-identity-drives-human [Hacker News] 370 points · 190 comments Hacker News Thu, 01 Oct 2026 20:02:57 +0000 intelsi.si:hn-49926411 https://huggingface.co/papers/2610.03391 [HF Daily Papers] 73 upvotes HF Daily Papers Thu, 01 Oct 2026 20:00:00 +0000 intelsi.si:hfp-2610.03391 https://huggingface.co/papers/2610.03574 [HF Daily Papers] 51 upvotes HF Daily Papers Thu, 01 Oct 2026 20:00:00 +0000 intelsi.si:hfp-2610.03574 https://huggingface.co/papers/2610.03665 [HF Daily Papers] 50 upvotes HF Daily Papers Thu, 01 Oct 2026 20:00:00 +0000 intelsi.si:hfp-2610.03665 https://openid.net/wp-content/uploads/2025/10/Identity-Management-for-Agentic-AI.pdf [Hacker News] 82 points · 28 comments Hacker News Thu, 01 Oct 2026 15:11:10 +0000 intelsi.si:hn-49922736 https://huggingface.co/papers/2610.01780 [HF Daily Papers] 250 upvotes HF Daily Papers Wed, 30 Sep 2026 20:00:00 +0000 intelsi.si:hfp-2610.01780 https://huggingface.co/papers/2610.02508 [HF Daily Papers] 74 upvotes HF Daily Papers Wed, 30 Sep 2026 20:00:00 +0000 intelsi.si:hfp-2610.02508 https://huggingface.co/papers/2610.02381 [HF Daily Papers] 53 upvotes HF Daily Papers Wed, 30 Sep 2026 20:00:00 +0000 intelsi.si:hfp-2610.02381 https://huggingface.co/papers/2610.02304 [HF Daily Papers] 45 upvotes HF Daily Papers Wed, 30 Sep 2026 20:00:00 +0000 intelsi.si:hfp-2610.02304 https://huggingface.co/papers/2610.00972 [HF Daily Papers] 44 upvotes HF Daily Papers Wed, 30 Sep 2026 20:00:00 +0000 intelsi.si:hfp-2610.00972