{"slug":"learning-adaptive","label":"Learning Adaptive","item_count":3,"day_count":3,"source_count":2,"first_seen":"2026-11-02T16:26:35+00:00","last_updated":"2026-11-09T18:58:41+00:00","generated_at":"2026-11-11T00:14:41.869679+00:00","sources":["arxiv_cs_ai","arxiv_cs_lg"],"days":[{"date":"2026-11-02","items":[{"title":"LAND: Learning an Adaptive Number of Draws for Bayesian Best-of-N Selection","url":"http://arxiv.org/abs/2611.03388v1","source":"arxiv_cs_ai","type":"paper","summary_1line":"Best-of-N decoding improves generation quality by sampling multiple candidates and selecting the best, but it allocates equal compute to sources regardless of how much they benefit from additional sampling: the expected benefit is small for sources with low-variance score distrib","sid":"72c9dee23b49167c","published":"2026-11-02T16:26:35+00:00","editor_note":"Inference-side entry: per-input sample counts instead of a fixed N."}]},{"date":"2026-11-06","items":[{"title":"Adaptive Scheduling Pipeline for Multi-Instance Asynchronous Reinforcement Learning","url":"http://arxiv.org/abs/2611.08864v1","source":"arxiv_cs_lg","type":"paper","summary_1line":"We present the Adaptive Scheduling Pipeline (ASP), an offline rollout scheduler for multi-instance asynchronous reinforcement learning (RL) of large language models (LLMs). ASP jointly reduces three sources of rollout underutilization - padding bubbles from intra-minibatch length","sid":"4f75e58fea1157aa","published":"2026-11-06T01:47:52+00:00","editor_note":"Training-side entry: rollout scheduling for asynchronous RL of LLMs."}]},{"date":"2026-11-09","items":[{"title":"Adaptive Quantization Preserves Fast Rates for TD Learning over Rate-Limited Channels","url":"http://arxiv.org/abs/2611.12878v1","source":"arxiv_cs_lg","type":"paper","summary_1line":"Motivated by reinforcement learning subject to communication constraints, we study a policy evaluation problem with linear function approximation over a rate-limited channel. In our setup, an agent interacting with an...","why_it_matters":"Matches feed focus: agent, evaluation.","sid":"ad7659407811cbec","published":"2026-11-09T18:58:41+00:00","editor_note":"Theory entry: convergence guarantee under a communication budget."}]}],"editorial":{"tldr":"This is not one developing event: three unrelated arXiv papers share the word \"adaptive\" and one idea, spending scarce resources where they help most. LAND (Nov 2) varies Best-of-N draws per input, and ASP (Nov 6) schedules RL rollouts to cut idle GPU time.","stale":false,"whats_new":"A Nov 9 theory paper shows adaptive quantization keeps TD learning's fast convergence rate even when an agent's updates cross a rate-limited channel.","why_it_matters":"Each paper targets a different budget: inference samples (LAND), RL-training GPU utilization (ASP), and communication bandwidth (TD quantization); only the first two touch production LLM stacks today.","take_for_builders":"If you run Best-of-N or async RL, read LAND and ASP for per-input sample budgets and rollout scheduling; treat the TD paper as theory, not a deployment change.","beats":[{"kicker":"INFERENCE","tone":"launch","headline":"LAND learns how many Best-of-N draws each input deserves","summary":"Fixed-N sampling wastes compute on inputs whose candidate scores barely vary.","sids":["72c9dee23b49167c"]},{"kicker":"TRAINING","tone":"rising","headline":"ASP schedules asynchronous RL rollouts to reduce underutilization","summary":"Targets padding bubbles and other idle time across multi-instance LLM RL rollouts.","sids":["4f75e58fea1157aa"]},{"kicker":"THEORY","tone":"neutral","headline":"Adaptive quantization preserves TD-learning rates over rate-limited links","summary":"Policy evaluation with linear function approximation under a bandwidth constraint.","sids":["ad7659407811cbec"]}],"open_questions":["Does LAND publish compute savings against fixed-N Best-of-N on a standard reward-model benchmark?","Is ASP's scheduler released as code, or integrated into an open RL framework?","Does the adaptive-quantization result extend beyond linear function approximation?"],"generated_at":"2026-11-11T00:13:33Z"}}