LoCo compresses LLM reasoning chains by 41.7-45.4% on math tasks
Original title: Think Less, Locally: Local Preference Learning for Reasoning Compression
LoCo trains models to stop overthinking using local preference learning on self-generated traces.
Large reasoning models often continue deliberating after sufficient information is available. LoCo detects overthinking states from model traces, freezes corresponding prefixes, regenerates shorter verified continuations, and trains with DPO on state-conditioned preference pairs. This teaches the model when to conclude rather than always shortening trajectories. On five mathematical and three code generation benchmarks, LoCo reduces completion length by 41.7-45.4% on math and 23.4-36.3% on code while maintaining or improving accuracy.
- reduces length by 41.7%-45.4% on mathematics
- reduces by 23.4%-36.3% on code generation
- maintains or improves accuracy
- best AUC across all code benchmarks
- cross-domain transfer from math-only training
- Organizations
- Hugging Face
- AI models
- large reasoning models
From the original article
1 min read at huggingface.co
Large reasoning models achieve strong performance on challenging tasks, but their long chain-of-thought often incurs substantial inference costs by continuing to deliberate after sufficient information has been obtained. Existing reasoning compression methods primarily operate on reasoning trajectories, paying less attention to the local decision of whether further reasoning is necessary given the current state.
By Muyao Yuan, Liyan Xu, Weizhan Zhang · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.