Share More, Search Less: Collaborative Parallel Thinking for Efficient Test-Time Scaling

Xinglin Wang, Hao Lin, Shaoxiong Feng, Peiwen Yuan, Yiwei Li, Jiayi Shi, Yueqi Zhang, Chuyi Tan, Ji Zhang, Boyuan Pan, Yao Hu, Kan Li

arXiv:2605.27030 · 2026-05-27 공개 · arXiv · PDF

llm test-time-scaling search-optimization latency-accuracy-tradeoff aime-benchmark inference-framework parallel-branching collaborative-parallel-thinking

Abstract

Test-Time Scaling (TTS) enhances the reasoning capabilities of large language models by allocating additional inference compute to explore the solution space. However, existing parallel TTS methods typically keep branches isolated during search: intermediate discoveries remain branch-private and cannot guide other branches in time. This information isolation causes substantial redundant exploration, as branches repeatedly rediscover information already found elsewhere and require more search steps to collect complete decision information needed to reach correct answers. To bridge this gap, we propose Collaborative Parallel Thinking (CPT), a training-free inference framework that enables search-time information sharing across parallel branches. CPT extracts compact intermediate information from ongoing branches, maintains a deduplicated query-level information pool, and broadcasts pool entries through the input context, allowing each branch in subsequent search steps to reuse discoveries made by other branches rather than rediscover the same information. Empirically, experiments on HMMT and AIME benchmarks show that CPT establishes a stronger accuracy--latency Pareto frontier than strong baselines across rollout budgets and model scales, highlighting search-time collaboration as an effective direction for efficient parallel TTS.

한국어 요약

📋 한 줄 요약

**[Test-Time Scaling / Parallel Thinking]** CPT가 parallel TTS branch 간 compact intermediate information을 deduplicated pool로 공유·broadcast해 redundant exploration 제거, HMMT·AIME에서 강한 accuracy-latency Pareto frontier.

🎯 핵심 기여도

💡 핵심 아이디어

Parallel TTS의 효율성을 제약하는 근본 원인은 branch 간 정보 격리이며, training-free하게 compact intermediate information을 추출·deduplicate·broadcast하는 search-time collaboration이 redundant exploration을 제거해 accuracy-latency Pareto frontier를 push한다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: Parallel TTS의 정보 격리 문제를 명확 진단하고 training-free 해법 제시, AIME·HMMT 등 수학 reasoning 벤치마크에서 정량 검증, 기존 TTS 방법과 호환되는 plug-in 가능성. **한계**: Information extraction·deduplication 품질이 효과 좌우, 매우 긴 context에서 pool entry broadcast의 token 비용 증가 가능, 수학 reasoning 중심 평가로 다른 도메인 일반화는 추가 검증.

🚀 실용적 활용