{"slug":"routing-learning","label":"Routing Learning","item_count":3,"day_count":3,"source_count":2,"first_seen":"2026-10-20T09:49:56+00:00","last_updated":"2026-11-04T18:10:33+00:00","generated_at":"2026-11-06T00:13:28.162297+00:00","sources":["arxiv_cs_ai","arxiv_cs_lg"],"days":[{"date":"2026-10-20","items":[{"title":"APRIL: Ambiguity-Aware Expert Routing for Respiratory Sound Domain-Incremental Learning","url":"http://arxiv.org/abs/2610.27942v1","source":"arxiv_cs_ai","type":"paper","summary_1line":"Automated respiratory sound classification enables continuous, non-invasive respiratory assessment, yet long-term deployment remains challenging due to domain shifts from evolving recording devices and acquisition per...","why_it_matters":"Matches feed focus: eval.","sid":"4c39e83baea4390b","published":"2026-10-20T09:49:56+00:00","editor_note":"Medical-audio expert routing under device domain shift; the thread's origin by keyword only."}]},{"date":"2026-10-30","items":[{"title":"TalentRL: Strategy-Guided Reinforcement Learning for Hierarchical Skill Routing","url":"http://arxiv.org/abs/2610.41143v1","source":"arxiv_cs_lg","type":"paper","summary_1line":"Natural-language skills enable large language model (LLM) agents to reuse procedures across tasks, yet most systems retrieve individual skills from a flat SkillBank without explicitly modeling the high-level strategy...","why_it_matters":"Matches feed focus: agentic, eval.","sid":"29493d021878283c","published":"2026-10-30T02:03:55+00:00","editor_note":"Moves routing to choosing skills from a SkillBank via a high-level strategy."}]},{"date":"2026-11-04","items":[{"title":"Visual Token Routing: Learning to keep tokens that matter","url":"http://arxiv.org/abs/2611.06854v1","source":"arxiv_cs_lg","type":"paper","summary_1line":"Computer-use agents (CUAs) have recently achieved superhuman performance on several long-horizon benchmarks. However, their computational cost limits their use in long-horizon tasks and real-time domains such as video...","why_it_matters":"Matches feed focus: agent.","sid":"9e076e903f5a5a8f","published":"2026-11-04T18:10:33+00:00","editor_note":"Applies learned routing to visual tokens to cut computer-use agent cost."}]}],"editorial":{"tldr":"Three unrelated arXiv papers share the word routing: expert routing for respiratory-sound classifiers under device drift (Oct 20), then strategy-guided RL for picking agent skills (Oct 30). The link is vocabulary, not a continuing research line.","stale":false,"whats_new":"A Nov 4 arXiv paper learns which visual tokens a computer-use agent keeps, targeting the compute cost of long-horizon and real-time agent tasks.","why_it_matters":"Learned routing is showing up in agent cost control (skills, tokens), but these are single-paper claims with no shared benchmark or released tooling.","take_for_builders":"Treat these as ideas to watch: if you run computer-use agents, wait for reproducible cost numbers before changing token or skill selection.","status":{"state":"Loose thread","tone":"neutral","changed":"2026-11-04","detail":"Three papers from two arXiv feeds reuse the routing idea in different domains; no cross-citation or common evaluation is shown."},"beats":[{"kicker":"EARLY","tone":"launch","headline":"Expert routing for respiratory-sound domain-incremental learning","sids":["4c39e83baea4390b"]},{"kicker":"AGENTS","tone":"rising","headline":"TalentRL applies strategy-guided RL to hierarchical skill routing for LLM agents","sids":["29493d021878283c"]},{"kicker":"NOW","tone":"now","headline":"Visual Token Routing learns which tokens computer-use agents keep","sids":["9e076e903f5a5a8f"]}],"open_questions":["Do TalentRL or Visual Token Routing release code or report results on a public agent benchmark?","Does the token-routing paper report a measured compute or latency reduction for computer-use agents?"],"generated_at":"2026-11-06T00:12:54+00:00"}}