Research

Paper

AI LLM February 23, 2026

Ada-RS: Adaptive Rejection Sampling for Selective Thinking

Authors

Yirou Ge, Yixi Li, Alec Chiu, Shivani Shekhar, Zijie Pan, Avinash Thangali, Yun-Shiuan Chuang, Chaitanya Kulkarni, Uma Kona, Linsey Pang, Prakhar Mehrotra

Abstract

Large language models (LLMs) are increasingly being deployed in cost and latency-sensitive settings. While chain-of-thought improves reasoning, it can waste tokens on simple requests. We study selective thinking for tool-using LLMs and introduce Adaptive Rejection Sampling (Ada-RS), an algorithm-agnostic sample filtering framework for learning selective and efficient reasoning. For each given context, Ada-RS scores multiple sampled completions with an adaptive length-penalized reward then applies stochastic rejection sampling to retain only high-reward candidates (or preference pairs) for downstream optimization. We demonstrate how Ada-RS plugs into both preference pair (e.g. DPO) or grouped policy optimization strategies (e.g. DAPO). Using Qwen3-8B with LoRA on a synthetic tool call-oriented e-commerce benchmark, Ada-RS improves the accuracy-efficiency frontier over standard algorithms by reducing average output tokens by up to 80% and reducing thinking rate by up to 95% while maintaining or improving tool call accuracy. These results highlight that training-signal selection is a powerful lever for efficient reasoning in latency-sensitive deployments.

Metadata

arXiv ID: 2602.19519
Provider: ARXIV
Primary Category: cs.AI
Published: 2026-02-23
Fetched: 2026-02-24 04:38

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2602.19519v1</id>\n    <title>Ada-RS: Adaptive Rejection Sampling for Selective Thinking</title>\n    <updated>2026-02-23T05:20:15Z</updated>\n    <link href='https://arxiv.org/abs/2602.19519v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2602.19519v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Large language models (LLMs) are increasingly being deployed in cost and latency-sensitive settings. While chain-of-thought improves reasoning, it can waste tokens on simple requests. We study selective thinking for tool-using LLMs and introduce Adaptive Rejection Sampling (Ada-RS), an algorithm-agnostic sample filtering framework for learning selective and efficient reasoning. For each given context, Ada-RS scores multiple sampled completions with an adaptive length-penalized reward then applies stochastic rejection sampling to retain only high-reward candidates (or preference pairs) for downstream optimization. We demonstrate how Ada-RS plugs into both preference pair (e.g. DPO) or grouped policy optimization strategies (e.g. DAPO). Using Qwen3-8B with LoRA on a synthetic tool call-oriented e-commerce benchmark, Ada-RS improves the accuracy-efficiency frontier over standard algorithms by reducing average output tokens by up to 80% and reducing thinking rate by up to 95% while maintaining or improving tool call accuracy. These results highlight that training-signal selection is a powerful lever for efficient reasoning in latency-sensitive deployments.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <published>2026-02-23T05:20:15Z</published>\n    <arxiv:primary_category term='cs.AI'/>\n    <author>\n      <name>Yirou Ge</name>\n    </author>\n    <author>\n      <name>Yixi Li</name>\n    </author>\n    <author>\n      <name>Alec Chiu</name>\n    </author>\n    <author>\n      <name>Shivani Shekhar</name>\n    </author>\n    <author>\n      <name>Zijie Pan</name>\n    </author>\n    <author>\n      <name>Avinash Thangali</name>\n    </author>\n    <author>\n      <name>Yun-Shiuan Chuang</name>\n    </author>\n    <author>\n      <name>Chaitanya Kulkarni</name>\n    </author>\n    <author>\n      <name>Uma Kona</name>\n    </author>\n    <author>\n      <name>Linsey Pang</name>\n    </author>\n    <author>\n      <name>Prakhar Mehrotra</name>\n    </author>\n  </entry>"
}