Latest

Every published item, in order · 7652 items
Oct 16
PapersBuildersMulti-sourceSelected09:41
System1-Arena benchmark tests which decisions LLMs can make without reasoning

The benchmark gives a concrete cost-accuracy-latency trade-off table for choosing between cheap probability-returning models and expensive reasoning models in production decision pipelines. Teams building flagging, escalation or agent control systems can use the 105-task and 138-environment breakdowns to match model class to task type. The finding that open LLMs beat System-1 models in closed loops challenges assumptions about small fast models in agentic settings.

Event
arXiv: OpenAI@Tianyu Chen et al. (5)Source
Tomorrow 08:00, the daily lands in your inboxTelegram channel or email. Same screening as this page, no ads.
PapersBuildersSelected09:29
ChainGraph: Neo4j-Based Exploit-Chain Reasoning for Penetration-Testing Agents

The result suggests the weakness in agentic security testing lies in flat context windows rather than model capability, so externalizing memory to a graph is a fix practitioners can apply today. Security engineers and red-team builders can wire Neo4j into existing MCP-compatible agents like Claude Code and Codex without custom tooling. The findings are illustrative single-run measurements, so treat them as a design pattern to replicate rather than a benchmark.

arXiv: OpenAI@Joas Antonio dos Santos BarbosaSource
PapersBuildersSelected09:28
IneqAgent Pairs LLM Agents with Symbolic Backends for Lean4 Inequality Proving

The interesting part is the division of labour: the LLM handles high-level lemma strategy while exact symbolic backends supply certificates, keeping every step Lean-verified and cheap at low inference budgets. Formal-methods researchers and theorem-proving engineers can reuse the MCTS-plus-backend pattern for other proof domains. It also gives DeepSeek-V4-Flash a concrete reasoning benchmark data point against other backbones tested in the paper.

arXiv: DeepSeek@Gaolei He et al. (5)Source
Oct 15

Only the latest 2000 items are listed; for older ones see the daily archive