Representation Consistency for Accurate and Coherent LLM Answer Aggregation
This work addresses the challenge of achieving accurate and coherent answer aggregation in LLMs during inference, offering a method that is incremental but effective for enhancing reasoning tasks.
The paper tackled the problem of improving large language models' performance via test-time scaling by introducing representation consistency (RC), a method that aggregates answers from multiple candidate responses by considering both answer frequency and the consistency of internal activations, resulting in consistent accuracy improvements of up to 4% over baselines on reasoning datasets.
Test-time scaling improves large language models' (LLMs) performance by allocating more compute budget during inference. To achieve this, existing methods often require intricate modifications to prompting and sampling strategies. In this work, we introduce representation consistency (RC), a test-time scaling method for aggregating answers drawn from multiple candidate responses of an LLM regardless of how they were generated, including variations in prompt phrasing and sampling strategy. RC enhances answer aggregation by not only considering the number of occurrences of each answer in the candidate response set, but also the consistency of the model's internal activations while generating the set of responses leading to each answer. These activations can be either dense (raw model activations) or sparse (encoded via pretrained sparse autoencoders). Our rationale is that if the model's representations of multiple responses converging on the same answer are highly variable, this answer is more likely to be the result of incoherent reasoning and should be down-weighted during aggregation. Importantly, our method only uses cached activations and lightweight similarity computations and requires no additional model queries. Through experiments with four open-source LLMs and four reasoning datasets, we validate the effectiveness of RC for improving task performance during inference, with consistent accuracy improvements (up to 4%) over strong test-time scaling baselines. We also show that consistency in the sparse activation signals aligns well with the common notion of coherent reasoning.