Vision models split objects from abstract relations
A preprint finds an early object-matching circuit and a later relation-matching circuit in vision-language models, then links the latter to performance on ARC-AGI-1.
A preprint finds an early object-matching circuit and a later relation-matching circuit in vision-language models, then links the latter to performance on ARC-AGI-1.
The Apache-licensed project publishes its accelerator design, compiler, simulator and host tools, with measured throughput showing a small, memory-bound system rather than a GPU replacement.
A preregistered experiment compares nine Claude Code skills with equally long neutral instructions, finding two cheaper than placebo, one worse and six statistically indistinguishable.
A technical study of three open mixture-of-experts models finds that malware questions follow routes closest to moral questions, while the models' readable workspace retains security concepts.
Research, model releases, local projects, technical essays and community reports: 51 concise items with direct sources.
The 501-billion-parameter mixture-of-experts model is available only through early access. Reflection says weights, a technical report and developer tools will follow later in October.
A preprint finds a shared internal direction that points language models to the first, second or later fact in a passage. Moving a question along that direction can change which fact the model retrieves.
A preprint finds that much of the benefit from an agent's history survives after its past actions are shuffled. Explicitly pairing each action with its result improves task completion.
The open-source command-line tool embeds llama.cpp and ships two small fine-tuned models. Its author also released the 401,975-pair training set and benchmark code.
Context Language Models replace an append-only transcript with a file the model can rewrite, delete and reorder. The authors report better long-task accuracy with less repeated computation after training the editing policy.
A practical guide measures the trade-off in separating prompt processing from token generation. Tail latency improves under load, but moving the model's working memory delays the first token.
Today's remaining AI news covers an unusual hybrid model, spatial-memory and honesty studies, community measurements, agent security incidents, EU watermarking and practical talks.