https://www.seventnews.com/en/articles/every-top-linear-recommender-may-reduce-to-one-of-two-regularizers SevenTnewS en 2026-09-20T20:00:00+02:00 Every top linear recommender may reduce to one of two regularizers https://www.seventnews.com/en/articles/explainability-assistant-parses-94-of-queries-beating-talktomodels-768 SevenTnewS en 2026-09-20T16:00:00+02:00 Explainability Assistant parses 94% of queries, beating TalkToModel's 76.8% https://www.seventnews.com/en/articles/a-33-of-36-result-published-with-the-authors-own-asterisk SevenTnewS en 2026-09-20T12:00:00+02:00 A 33-of-36 result, published with the authors' own asterisk https://www.seventnews.com/en/articles/qwen-llama-and-gemma-split-on-when-a-model-stops-routing-and-starts-answering SevenTnewS en 2026-09-20T08:00:00+02:00 Qwen, Llama and Gemma split on when a model stops routing and starts answering https://www.seventnews.com/en/articles/sirfs-151-point-gain-lives-in-the-weights-not-the-prompt SevenTnewS en 2026-09-19T20:00:00+02:00 SIRF's 15.1-point gain lives in the weights, not the prompt https://www.seventnews.com/en/articles/a-panel-of-ai-agents-can-agree-and-still-be-wrong-paper-argues SevenTnewS en 2026-09-19T16:00:00+02:00 A panel of AI agents can agree and still be wrong, paper argues https://www.seventnews.com/en/articles/cobra-skills-cuts-agent-skill-tuning-costs-by-55-58-on-50-examples SevenTnewS en 2026-09-19T12:00:00+02:00 COBRA-Skills cuts agent skill tuning costs by 55-58% on 50 examples https://www.seventnews.com/en/articles/mistrals-40000-line-fortran-rewrite-shows-where-agent-autonomy-runs-out SevenTnewS en 2026-09-19T08:00:00+02:00 Mistral's 40,000-line Fortran rewrite shows where agent autonomy runs out