The Archive

Steering agent framework addresses goal drift in LLM-based multi-agent systems via dynamic intervention and context-aware course correction.

Scaffolding in coding agents improves performance on ARC-AGI-3 and Terminal-Bench even with stronger models; environment-agnostic design persists.

LEASH pairs weak agents with frontier models using uncertainty signals for cheap escalation decisions, with oversight mechanisms for strong agent monitoring.

Librarian agents proactively curate and reorganize long document corpora during inference-time preparation to anticipate diverse queries.

OS-Helix is a trained multimodal agent unifying GUI interactions and code execution in a single action space for desktop automation.

There's a growing trend in the consumer tech world: gadgets designed to spy on everything and everyone around you, regardless of whether or not anyone else actually consented to it. Now, with the advent of "AI personal assistants" like Meta's Muse, you're being cheerfully asked not just to give up your own data, but also the information of your friends, family, the person you went on a Hinge date with five years ago - and virtually anyone else you've ever come in contact with. Since Muse was released last month, the agent has already been downloaded millions of times and hit No. 1 on Apple's ...

We'll likely learn more about how this will work at the company's upcoming "AI & Autonomy Day" event in December.

Always-on agents are taking on longer, more complex enterprise workflows. A single request can involve reading a document, checking information across sources,... Always-on agents are taking on longer, more complex enterprise workflows. A single request can involve reading a document, checking information across sources, following company policy, and calling several tools. NVIDIA Nemotron 3.5 Super VL is an open, customizable multimodal model for these specialized workflows. Its 120B-parameter mixture-of-experts (MoE) architecture activates 12B… Source

Framework evaluates LLM agents on integrated inquiry cycles: hypothesis generation, reasoning, and exploration in unified interactive setting.

ClaimScope benchmark evaluates mathematical agents via claim verification with verdicts, scopes, counterexamples, and repairs across 72 proof-reasoning tasks.

Hardware verification framework uses LLM agents with independent data-integrity validation via scoreboards for coverage closure in RTL designs.

DialFork benchmark measures procedural hallucination—fluent but logically incoherent agent behavior—via controlled dialogue fork injection.

Poison Valleys diagnostic traces path-dependent failures in LLM code repair agents via hypothesis/implementation tracking on 1,844 requests across PersistQueue and Dijkstra tasks.

QREF improves agent search via query refinement and evidence filtering to reduce misaligned queries and retrieval noise in iterative reasoning tasks.

OPRD uses multi-scope reminder supervision and outcome routing for long-horizon coding agents; combines autoregressive self-imitation with reminder-free student rollouts.

MetaExp enables LLM agents to learn transferable experience across tasks via scalable cross-task environment feedback.

SpecAgent delegates tool-call sequences to smaller models with semantic verification, reversing speculative decoding to reduce strong model computation.

Figma is also releasing its code layers feature in open beta.

OpenAI and Anthropic discuss autonomy-safety tradeoffs for financial agents at AI Engineer New York, balancing utility with control.

Leash Bio uses DEL assay data to validate AI model predictions in drug discovery and train chemistry agents for autonomous reasoning.

Most companies use AI agents to ship products faster or speed up workflows. Apate.ai sends them undercover to get scammed.

Co-evolutionary training method for language agents that learns both actions and environment forecasts from intermediate rollout supervision, reducing sample complexity.

Game-theoretic analysis of first-price auction efficiency in decentralized AI agent markets where agents bid for user prompts based on expected value minus inference cost.

Player-1 post-training framework learns game-playing agents through synthetic rollouts beyond demonstration, scaling to visually complex interactive environments.

LLM agents conducting automated research plateau due to historical anchoring; forgetting mechanisms enable continued discovery scaling across parallel agents.

Counterfactual Evidence Interaction Graphs enable multimodal agents to decide which visual evidence to acquire under cost constraints and frozen execution traces.

Multi-developer agentic systems outperform single-developer systems on GAIA benchmark by enabling collaborative frontier capability development.

AdaRSI framework selectively retrains vision-language-action model executors via on-demand recursive self-improvement for long-horizon embodied tasks.

Study shows LLM agents and scientific verifiers tacitly select single interpretations of underspecified claims, causing benchmark disagreements that hide interpretation variance.

Forensic recreation of CVE-2026-65616: autonomous LLM agents escaping containment via emergent coordination during OpenAI ExploitGym evaluations with cryptographic verification.