paper-with-me

Papers

SABER: Switchable and Balanced Training for Efficient LLM Reasoning

2025-08-08 · Kai Zhao, Yanjun Zhao, Jiaming Song, Shien He, Lusheng Zhang, Qiang Zhang, Tianjiao Li arxiv

Large language models (LLMs) empowered by chain-of-thought reasoning have achieved impressive accuracy on complex tasks but suffer from excessive inference costs and latency when applied uniformly to all problems. We propose SABER (Switchable and Balanced Training for Efficient LLM Reasoning), a reinforcement learning framework that endows LLMs with user-controllable, token-budgeted reasoning. SABER first profiles each training example's base-model thinking token usage and assigns it to one of the predefined budget tiers. During fine-tuning, the model is guided by system prompts and length-aware rewards to respect its assigned budget. In parallel, we incorporate no-think examples to ensure the model remains reliable even when explicit reasoning is turned off. SABER further supports four discrete inference modes - NoThink, FastThink, CoreThink, and DeepThink, enabling flexible trade-offs between latency and reasoning depth. Extensive evaluations on math reasoning (MATH, GSM8K), code generation (MBPP), and logical reasoning (LiveBench-Reasoning) demonstrate that SABER achieves high accuracy under tight budgets, graceful degradation, and effective cross-scale and cross-domain generalization. In particular, SABER-FastThink cuts reasoning length by 65.4% and yields a 3.6% accuracy gain compared with the base model on the MATH benchmark.

📄 PDF Abstract BibTeX arXiv:2508.10026

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement LearningDomain GeneralizationLogical ReasoningCode Generation

Similar Papers 제목 키워드 기반

SABER: Stability-Aware Early Exit for LLM Reasoning via Adversarial Branch Probing

2026-08-28 · Wanli Cheng, Haiya Xiang, Juntao Li, Hongling Wang 외 arxiv

Large Reasoning Models (LRMs) achieve strong reasoning capabilities, yet long-chain reasoning becomes inefficient once the intermediate answer stabilizes across reasoning steps: additional reasoning yields little margina…

Integrating Fine-Grained Audio-Visual Evidence for Robust Multimodal Emotion Reasoning

2026-01-26 · Zhixian Zhao, Wenjie Tian, Lei Xie arxiv

Multimodal emotion analysis is shifting from static classification to generative reasoning. Beyond simple label prediction, robust affective reasoning must synthesize fine-grained signals such as facial micro-expressions…

Multimodal Reasoning

Trust or Abstain? A Self-Aware RAG Approach

2026-05-11 · Xi Zhu, Ziqi Wang, Kai Mei, Wujiang Xu 외 arxiv

Retrieval-augmented generation (RAG) improves large language models (LLMs) by incorporating external evidence, but it also introduces knowledge conflicts when retrieved contextual knowledge (CK) and parametric knowledge …

Investigating associative, switchable and negatable Winograd items on renewed French data sets

2022-06-01 · JEP/TALN/RECITAL 2022 6 · Xiaoou Wang, Olga Seminck, Pascal Amsili

The Winograd Schema Challenge (WSC) consists of a set of anaphora resolution problems resolvable only by reasoning about world knowledge. This article describes the update of the existing French data set and the creation…

NegationWorld Knowledge

SABER: A Scalable Action-Based Embodied Dataset for Real-World VLA Adaptation

2026-05-10 · Narsimha Menga, Parikshit Sakurikar, Amirreza Rouhi, Satya Sai Reddy 외 arxiv

Robotic deployment in real-world environments depends on rich, domain-specific action data as much as on strong model architecture. General-purpose robot foundation models show modest performance in complex unseen tasks …