The paper investigates test-time scaling for small-scale open-ended reasoning models by introducing semantic cluster voting, a judge-free method that groups model outputs by meaning and selects the largest cluster’s representative. Experiments on a 250-prompt benchmark across three 3‑B models show that voting improves accuracy on math tasks (10–29 point gains) but only marginally on open-ended text (+0.30 points). A larger 200‑problem MATH‑500 run confirms a +8.5 point math improvement, while most open-ended samples fall into a single cluster, indicating limited diversity and that higher decoding temperatures reduce quality.
arXiv:2608.22048v1 Announce Type: new Abstract: Large language models are increasingly deployed on local hardware for privacy, cost, and accessibility reasons. Yet many evaluations emphasize accuracy...
arXiv:2606. 28661v1 Announce Type: cross Abstract: People overthink; language models over-sample, and the extra effort can talk both into a worse answer.
arXiv:2608. 05643v1 Announce Type: new Abstract: Test-time scaling improves LLM reasoning by using additional inference compute, but wider sampling alone can suffer from diminishing returns: new rollouts often repeat existing answer patterns instead of adding useful reasoning diversity.
arXiv:2607. 10139v1 Announce Type: cross Abstract: Selecting the correct answer from a pool of candidate reasoning chains is the engine of test-time scaling, yet the standard selectors each carry a cost: self-consistency inherits the errors of the single model it resamples, and trained reward models need labeled data and transfer poorly off-distribution.
arXiv:2608. 15445v1 Announce Type: new Abstract: When a reward is correct on every training example yet consistent with more than one goal, a model can acquire an unintended one, a failure known as goal misgeneralization.
arXiv:2607. 28576v1 Announce Type: cross Abstract: Methods that make a language model plan, criticise and rewrite its own answer, reflect on mistakes, pick the best of several attempts, or debate with copies of itself nearly all make it generate far more text than a single chain of thought.
The paper introduces Concise-9B, a version of the Qwen3.5-9B reasoning model that reduces unnecessary thinking tokens through concise self‑training. By applying two rounds of on‑policy rejection‑sampling fine‑tuning and a small share of answer‑synthesis targets, Concise-9B cuts token usage to roughly two‑thirds of the original model while maintaining comparable accuracy on a wide range of hard mathematics and reasoning benchmarks. The study also shows that most of the token savings occur after the answer has already formed, and that the approach’s effectiveness depends on the verbosity of the base model.
The paper investigates how test‑time computation can enhance language models and at what cost, introducing the SELF‑POT benchmark to evaluate this across competition mathematics, competitive programming, and agentic workflows. SELF‑POT separates candidate coverage from final accuracy, tracks correctness transitions under revision, and measures protocol completion alongside task success. Using a unified budget rule, the study compares direct inference, parallel sampling, and self‑revision across five low‑cost reasoning models, revealing that selection rules and failure handling significantly influence gains and cost savings.
The paper examines test‑time scaling (TTS) methods that use extra inference compute to improve language model outputs. Across five open‑ended benchmarks—medicine, law, finance, general chat, and creative writing—the study finds that increasing exploration (generating more candidates) consistently yields better top candidates, but exploitation (selecting the final output) remains weak due to poor reward‑model correlation. Only the Fusion approach, which synthesizes candidates, reliably improves results, yet it recovers only about 40% of the potential quality, indicating that the bottleneck lies in choosing from the candidate pool rather than generating it.
The paper investigates test‑time scaling for language models, noting that sequential methods often stall due to entering attractors—sets of answers that block further exploration. Across 27 experiments, over half of sequential trajectories hit an attractor within four iterations. Introducing a simple model‑mixing intervention reduces attractor incidence by 21.2 percentage points, expands solution coverage beyond a compute‑matched parallel baseline, and boosts recursive self‑aggregation accuracy by at least 2.2 percentage points.
The paper introduces SPLIT, a method that optimally allocates test‑time compute among sampling, model selection, and verification for language models and agents. Using an exact probability formula that accounts for task difficulty, shared blind spots, and verifier correlations, SPLIT is fitted via empirical Bayes on a small labeled pilot and then used to plan budgets for unseen tasks. Across seven diverse benchmarks, 26 models, and six verifier types, SPLIT achieves near‑optimal performance, outperforming rule‑based baselines by up to 2.5 percentage points and providing insights into when additional LLM judges reduce precision.
arXiv:2510. 00492v3 Announce Type: replace Abstract: The reliability of large language models (LLMs) during test-time scaling is often assessed with \emph{external verifiers} or \emph{reward models} that distinguish correct reasoning from flawed logic.