@inproceedings{ge-etal-2026-ada,
title = "{A}da-{RS}: Adaptive Rejection Sampling for Selective Thinking",
author = "Ge, Yirou and
Li, Yixi and
Chiu, Alec M. and
Shekhar, Shivani and
Pan, Zijie and
Thangali, Avinash and
Chuang, Yun-Shiuan and
Kulkarni, Chaitanya and
Kona, Uma and
Pang, Linsey and
Mehrotra, Prakhar",
editor = "Li, Yunyao and
Rehm, Georg and
Tu, Mei",
booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 6: Industry Track)",
month = jul,
year = "2026",
address = "San Diego, California, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.acl-industry.88/",
doi = "10.18653/v1/2026.acl-industry.88",
pages = "1265--1276",
ISBN = "979-8-89176-394-4",
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 {\ensuremath{\sim}}80{\%} and reducing thinking rate by up to {\ensuremath{\sim}}95{\%} while maintaining or improving tool call accuracy. We further demonstrate that these gains generalize across model scales (Qwen3-1.7B, 8B, 14B) and domains ({\ensuremath{\tau}} 2-Bench airline and telecom). These results highlight that training signal selection is a powerful lever for efficient reasoning in latency-sensitive deployments."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ge-etal-2026-ada">
<titleInfo>
<title>Ada-RS: Adaptive Rejection Sampling for Selective Thinking</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yirou</namePart>
<namePart type="family">Ge</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yixi</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alec</namePart>
<namePart type="given">M</namePart>
<namePart type="family">Chiu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shivani</namePart>
<namePart type="family">Shekhar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zijie</namePart>
<namePart type="family">Pan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Avinash</namePart>
<namePart type="family">Thangali</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yun-Shiuan</namePart>
<namePart type="family">Chuang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chaitanya</namePart>
<namePart type="family">Kulkarni</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Uma</namePart>
<namePart type="family">Kona</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Linsey</namePart>
<namePart type="family">Pang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Prakhar</namePart>
<namePart type="family">Mehrotra</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yunyao</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Georg</namePart>
<namePart type="family">Rehm</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mei</namePart>
<namePart type="family">Tu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-394-4</identifier>
</relatedItem>
<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 \ensuremath\sim80% and reducing thinking rate by up to \ensuremath\sim95% while maintaining or improving tool call accuracy. We further demonstrate that these gains generalize across model scales (Qwen3-1.7B, 8B, 14B) and domains (\ensuremathτ 2-Bench airline and telecom). These results highlight that training signal selection is a powerful lever for efficient reasoning in latency-sensitive deployments.</abstract>
<identifier type="citekey">ge-etal-2026-ada</identifier>
<identifier type="doi">10.18653/v1/2026.acl-industry.88</identifier>
<location>
<url>https://aclanthology.org/2026.acl-industry.88/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>1265</start>
<end>1276</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Ada-RS: Adaptive Rejection Sampling for Selective Thinking
%A Ge, Yirou
%A Li, Yixi
%A Chiu, Alec M.
%A Shekhar, Shivani
%A Pan, Zijie
%A Thangali, Avinash
%A Chuang, Yun-Shiuan
%A Kulkarni, Chaitanya
%A Kona, Uma
%A Pang, Linsey
%A Mehrotra, Prakhar
%Y Li, Yunyao
%Y Rehm, Georg
%Y Tu, Mei
%S Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, USA
%@ 979-8-89176-394-4
%F ge-etal-2026-ada
%X 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 \ensuremath\sim80% and reducing thinking rate by up to \ensuremath\sim95% while maintaining or improving tool call accuracy. We further demonstrate that these gains generalize across model scales (Qwen3-1.7B, 8B, 14B) and domains (\ensuremathτ 2-Bench airline and telecom). These results highlight that training signal selection is a powerful lever for efficient reasoning in latency-sensitive deployments.
%R 10.18653/v1/2026.acl-industry.88
%U https://aclanthology.org/2026.acl-industry.88/
%U https://doi.org/10.18653/v1/2026.acl-industry.88
%P 1265-1276
Markdown (Informal)
[Ada-RS: Adaptive Rejection Sampling for Selective Thinking](https://aclanthology.org/2026.acl-industry.88/) (Ge et al., ACL 2026)
ACL
- Yirou Ge, Yixi Li, Alec M. Chiu, Shivani Shekhar, Zijie Pan, Avinash Thangali, Yun-Shiuan Chuang, Chaitanya Kulkarni, Uma Kona, Linsey Pang, and Prakhar Mehrotra. 2026. Ada-RS: Adaptive Rejection Sampling for Selective Thinking. In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track), pages 1265–1276, San Diego, California, USA. Association for Computational Linguistics.