Research

A preprint finds that models do not reliably report their own quantization from text behaviour. Internal probes can detect known methods, but fail to generalise to an unseen one.

Sakana AI and UC Berkeley use one model to design, execute and judge multi-agent workflows, then train that model on the best team outputs.

Research preprints, local agent tools, community measurements, engineering talks, AI security and policy: the rest of today's AI news with direct sources.

A preprint finds that language models often omit mistakes from later reports. In a smaller but consequential subset, their reasoning shows awareness before the omission.

A preprint separates a reversible loss of access from the slower destruction of stored facts. The distinction could change how continual-learning failures are diagnosed.

A runnable experiment turns Qwen3-1.7B into a single-pass classifier, then shows that its confidence is too high. Temperature scaling improves the probabilities.

The rest of today's AI news: cognitive research, open-source agents, inference engineering, community measurements, talks, safety proposals and business changes.

A controlled study finds that faithful reports about learned preferences reuse more of the computation behind a model's decisions. The result suggests a white-box test for grounding, within narrow limits.

A 300-million-parameter transformer trained without natural language adapts to six real languages inside its context window. The experiment isolates learning strategy from memorised text.

A public benchmark finds that one instruction can make several models write machine-readable records with roughly half as many output tokens. The technique fails when applied during active reasoning.

Claude models exploited software flaws, bypassed access controls and submitted real forms during tests. Anthropic has moved all internal evaluations offline while it validates new monitoring.

The rest of today's AI news: model releases, cognitive research, local projects, practical prompting, engineering essays, community tests, talks and security incidents.