LATTS: Locally Adaptive Test-Time Scaling
This addresses the problem of inefficient resource use in test-time scaling for LLM applications, offering a more adaptive approach, though it is incremental as it builds on existing verifier-based methods.
The paper tackles the inefficiency of uniform test-time scaling in LLMs by proposing LATTS, which adaptively allocates compute per generation step based on local difficulty, achieving significantly better accuracy-compute tradeoffs than standard methods.
One common strategy for improving the performance of Large Language Models (LLMs) on downstream tasks involves using a \emph{verifier model} to either select the best answer from a pool of candidates or to steer the auto-regressive generation process towards better outputs. This class of methods typically results in improved accuracy at the cost of increased computation at test-time, a paradigm known as \emph{test-time scaling}. However, most existing approaches increase computation uniformly across all samples and generation steps, without considering the complexity of individual instances, leading to inefficient resource use. We address this limitation by proposing an approach, called \emph{Locally Adaptive Test-Time Scaling (LATTS)}, that allocates variable compute across generation steps. Specifically, at each generation step, LATTS employs a verifier-based acceptance criterion to decide whether to resample, backtrack, restart, or stop the generation process. This criterion effectively adjusts the per-step computational effort based on a precise notion of \emph{local difficulty} derived from the verifier model. Empirical results show that LATTS achieves significantly superior accuracy--compute tradeoffs compared to standard verifier-based methods.