We explore the role of verification granularity in sampling-based test-time scaling.

We find that excessively scaling Chain of Thought (CoT) length can impair the model's reasoning performance in certain domains, and we propose a Thinking-Optimal Scaling strategy to achieve more effective and efficient test-time scaling.

We propose DORA, a provably optimal resource allocation strategy for efficient test-time scaling.

We present an algorithm for test-time scaling of SDE-based diffusion models by searching for noise trajectories which optimize arbitrary rewards, empirically matching/exceeding MCTS performance.

We show that recurrent-depth transformers can be scaled to be effective language models, with particularly strong gains through additional compute for reasoning tasks.

We propose CADMorph, an inference-time editing method for parametric CAD models using the signal of geometry changes via a Plan–Generate–Verify Loop over pretrained priors, namely a LDM and an LLM, bypassing the need of non-existent editing data.

programming framework that does runtime compilation of an LLM-based agent workflow into a search space, enabling independent experimentation of different overlaying inference-time search strategies

demonstrates reasoning settings where scaling the length of chain-of-thought offers an exponential advantage over parallel scaling and aggregation.

We introduce the first theoretical framework for analyzing LLM reasoning errors, and bridge two typical sampling-based test-time scaling methods to achieve both low error and fast convergence.