Scalable LLM Math Reasoning Acceleration with Low-rank Distillation
Due to long generations, large language model (LLM) math reasoning demands significant computational resources and time. While many existing efficient inference methods have been developed with excellent performance preservation on language tasks, they often severely degrade math performance. In this paper, we propose Caprese, a low-cost distillation method to recover lost capabilities from deploying efficient inference methods, focused primarily in feedforward blocks. With original weights unperturbed, roughly 1% of additional parameters, and only 20K synthetic training samples, we are able to recover much if not all of the math capabilities lost from efficient inference for thinking LLMs and without harm to language tasks for instruct LLMs. Moreover, Caprese slashes the number of active parameters (~2B cut for Gemma 2 9B and Llama 3.1 8B) and integrates cleanly into existing model layers to reduce latency (>11% reduction to generate 2048 tokens with Qwen 2.5 14B) while encouraging response brevity.
Code (0)
등록된 구현이 없습니다.
Tasks
Language ModelingLanguage ModellingLarge Language ModelMathMethods 이 논문이 사용한 방법론
Similar Papers 제목 키워드 기반
LoRi: Low-Rank Distillation for Implicit Reasoning
Implicit chain-of-thought (iCoT) methods aim to internalize reasoning in large language models, but often underperform explicit CoT prompting. We empirically find that hidden-state reasoning trajectories exhibit low-rank…
Mathematical ReasoningReM-MoA: Reasoning Memory Sustains Mixture-of-Agents Scaling
Mixture-of-Agents (MoA) architectures improve inference-time scaling by organizing multiple LLM agents into layered reasoning pipelines. However, existing MoA variants fail to sustain gains as depth increases, exhibiting…
Formal LogicVOLD: Reasoning Transfer from LLMs to Vision-Language Models via On-Policy Distillation
Training vision-language models (VLMs) for complex reasoning remains a challenging task, i.a. due to the scarcity of high-quality image-text reasoning data. Conversely, text-based reasoning resources are abundant and sca…
Reinforcement LearningAnti-Self-Distillation for Reasoning RL via Pointwise Mutual Information
On-policy self-distillation, where a student is pulled toward a copy of itself conditioned on privileged context (e.g., a verified solution or feedback), offers a promising direction for advancing reasoning capability wi…
Distillation and Refinement of Reasoning in Small Language Models for Document Re-ranking
We present a novel approach for training small language models for reasoning-intensive document ranking that combines knowledge distillation with reinforcement learning optimization. While existing methods often rely on …
Document RankingInformation RetrievalKnowledge DistillationLanguage Modeling+4