HazardAuditor is an execution-grounded framework for training safety guard models for computer-use AI agents, addressing the gap between static prompt/response…
Meta engineers extended FlashAttention-4 with end-to-end MXFP8 block-scaled support for both forward and backward passes on NVIDIA Blackwell GPUs, addressing…
Researchers introduce Gavel, a skill-routing method that extracts selection signals directly from a frozen LLM’s mid-layer activations using only two…
The paper introduces Decoy Direction Optimization (DDO), a fast weight-editing defense against Refusal Feature Ablation attacks that bypass open-weight model…
Researchers built Emergence World, a continuously running multi-agent environment used to stress-test eight parallel 10-agent ‘worlds’ (seven single-model, one mixed-model)…
The paper introduces ImpossibleRubrics, a 169-task benchmark of ‘impossible’ prompts (each paired with an oracle certificate defining valid honest responses)…
The paper presents ZGCM-1, a fully open 7B dense foundation model trained from scratch to couple internal reasoning with external…
The paper proposes ModularRSI, a framework for evolving AI agent execution harnesses (coding/terminal agents) that decomposes the harness into five…
The paper studies ‘long-horizon memorization,’ where a model must learn 100 sequential query-answer tasks via continual fine-tuning without retaining past…
The post introduces Reef, an open-source infrastructure framework that treats an agent’s live inference server as the substrate for continuous…
The post describes a method for RL-training coding agents where the agent’s own harness owns its execution loop rather than…
Researchers identify four unbounded memory peaks in standard MoE training pipelines — expert dispatch, vocabulary projection, gradient checkpointing, and optimizer…
