2026-10-28

Successful tool-use trajectories are not necessarily useful supervision. Even successful interactions may underrepresent multi-step dependencies or contain actions that are unjustified by the information visible to the agent at decision time. We identify two coupled requirements for effective trajectory supervision: compositional coverage, so that important observation-action dependencies are represented, and decision grounding, so that each supervised action is supported by the learner-visible

🤖Trajectory2Gym improves tool-use training by constructing interactions with compositional coverage and decision grounding rather than collecting successful trajectories.

Continual learning is the ability of LLM agents to improve over time by acquiring, refining, and applying knowledge from experience while retaining useful prior knowledge. It is essential for agents learning over long time horizons, yet current evaluations largely assess individual task performance rather than improvement through accumulated experience. We introduce Continual Learning Exam (CLE), a large-scale benchmark developed by 50+ domain experts to measure experience-driven improvement in

🤖A new benchmark evaluates whether LLM agents improve through repeated professional tasks with feedback, revealing limited sustained learning in current frontier models.

Localized AI manipulations in long-form videos often affect brief events within otherwise authentic content, making video-level classification insufficient for forensic analysis. Existing audio-visual deepfake benchmarks largely focus on short talking-head clips, while long-form temporal forensics remains predominantly visual. We introduce AV-FABLE, a dual-timeline benchmark for sparse manipulations in open-scene long-form audio-visual videos. AV-FABLE contains 16,650 pristine--manipulated twin

🤖AV-FABLE introduces a long-form audio-visual benchmark for sparse manipulations, showing that frozen perception features of audio-visual large language models contain forensic evidence best recovered via separate temporal decoders rather than language outputs.

UC Berkeley’s LEANN vector index stores a pruned graph and recomputes embeddings at query time. It matches HNSW accuracy at 90% recall using 4 GB instead of 188 GB, at the cost of seconds-level search. The post Meet LEANN: An Open-Source Vector Index that Shrinks a 60M-Chunk RAG Index from 201 GB to 6 GB appeared first on MarkTechPost .

Owners of AI-run firms would evade liability for "reckless" AI, US Rep. Comer says.

A family is suing OpenAI, saying ChatGPT coached their daughter to quietly taper off her seizure medication and keep it from her doctors before she died in February.

The banking giant reminds customers that "you're responsible for activity" resulting from your AI agents' access.

You can use the AI assistant to update calendar events and search your inbox, among other things.

The deal, expected to close sometime next year, is to help bring AI agents to work alongside humans to underwrite, analyze, and price decisions.