From Long to Lean: Performance-aware and Adaptive Chain-of-Thought Compression via Multi-round Refinement
This addresses efficiency issues in large language models for users needing faster inference, though it appears incremental as it builds on existing CoT compression techniques.
The paper tackles the problem of Chain-of-Thought reasoning causing high inference latency due to verbosity, and proposes an adaptive compression method that improves accuracy by 5.6% over baselines while reducing CoT length by 47 tokens and lowering latency.
Chain-of-Thought (CoT) reasoning improves performance on complex tasks but introduces significant inference latency due to verbosity. We propose Multiround Adaptive Chain-of-Thought Compression (MACC), a framework that leverages the token elasticity phenomenon--where overly small token budgets can paradoxically increase output length--to progressively compress CoTs via multiround refinement. This adaptive strategy allows MACC to determine the optimal compression depth for each input. Our method achieves an average accuracy improvement of 5.6 percent over state-of-the-art baselines, while also reducing CoT length by an average of 47 tokens and significantly lowering latency. Furthermore, we show that test-time performance--accuracy and token length--can be reliably predicted using interpretable features like perplexity and compression rate on the training set. Evaluated across different models, our method enables efficient model selection and forecasting without repeated fine-tuning, demonstrating that CoT compression is both effective and predictable. Our code will be released in https://github.com/Leon221220/MACC.