automated math problem solvers

Abstract: Recent methods for Reinforcement Learning with Verifiable Rewards (RLVR) use rollout diversity to improve reasoning performance or training efficiency. Despite these gains, the extent to which differences among correct-answer rollouts reflect useful alternative reasoning or unresolved errors remains poorly understood. We investigate this question through a blind paired analysis of a model trained with Group Relative Policy Optimization (GRPO), comparing the highest- and lowest-agreement correct-answer traces for the same problem. Across 126 pairs, low-agreement traces have more spurious reasoning than high-agreement traces (35% vs. 22%), including internal inconsistencies and misuse of problem conditions. Based on these findings, we propose RoAR (Rollout-Agreement Reward), which adds reasoning-level agreement to GRPO's rewards for verified-correct rollouts. RoAR computes this signal from existing on-policy rollout groups, requiring no human-written rationales, additional sampling, or separately trained process evaluator. On five mathematical reasoning benchmarks, the 1.5B GRPO-RoAR model reaches 55.5% mean pass@1 under zero-shot evaluation using 7K training examples, exceeding GRPO and GRPO-CARE by 6.0 and 5.5 percentage points. A separate blind audit of 100 problems answered correctly by all four 1.5B models finds spurious rates of 18% for GRPO-RoAR, 32% for GRPO, 27% for GRPO-CARE, and 37% for DRA-GRPO. Our findings identify reasoning validity as an important consideration when using rollout differences to guide learning. Our code is available at https://github.com/ldi-kyunghee/RoAR.

Abstract: Self-evolving training improves the performance of agents on difficult tasks through verifiable feedback, but it often applies costly deliberation uniformly across tasks. To address this problem, we propose Mode-Optimized Reasoning with Self-Adaptive Effort (MORE), a self-evolving training framework that adaptively selects both reasoning form and computational effort for each task. Specifically, MORE follows a two-stage training paradigm: (i) \emph{Stage 1} establishes a compact prior over direct answering, short and long chain-of-thought, and code-based reasoning, and (ii) \emph{Stage 2} self-evolves this prior with Mode-Adaptive and Test-time Effort PPO (MATE-PPO) on generated, verifiable task streams without curated data. By learning mode selection and token expenditure jointly, MORE allocates concise reasoning to tasks that permit it while retaining extended deliberation where it is useful. Across coding, mathematics, and general-reasoning benchmarks, MORE achieves consistent performance gains on Base and Coder backbones at model scales from 3B to 14B. During inference, MORE demonstrates higher accuracy and better efficiency, with markedly shorter response lengths and reduced inference time.

Abstract: Best-of-N methods improve text-only LLM reasoning by sampling multiple trajectories before selecting an answer. For Vision-Language Models (VLMs), answer frequency and model confidence alone can be insufficient, as sampled generations also vary in their engagement with the visual input. We find that distributional confidence and visual attention are complementary indicators of rollout quality. Each is limited on its own, while trajectories that show strong signals in both are more likely to be correct. Motivated by this, we introduce VisConf, a rollout-quality score that multiplicatively combines visual engagement with candidate-calibrated Self-Certainty, penalizing rollouts that are weak in either signal. We incorporate Rank-Weighted Consensus (RWC), a tuning-free aggregation rule that converts within-question VisConf ranks into answer-level vote weights. Across three VLMs on MathVista, MMMU-Pro, and MMStar, VisConf with RWC consistently outperforms strong BoN baselines. Our analysis shows particularly large gains when correct trajectories are present but underrepresented in the candidate pool.

Abstract: Chain-of-thought (CoT) improves reasoning but incurs substantial autoregressive rationale cost. Continuous CoT moves intermediate computation into recurrent hidden states, yet typically uses a fixed number of updates even though the best budget varies across questions and accuracy need not improve monotonically with $K$. We introduce Dynamic K-Head, a question-conditioned, pre-rollout budget selector. It is jointly trained with the reasoner under selected budgets, using difficulty supervision based on scores assigned from the question alone. We also show that latent-update count can misrepresent end-to-end cost because a short latent rollout can shift computation into a longer explicit continuation. We therefore count latent updates, implementation-specific special tokens, and all decoded text. Instantiated with CoDI on Llama 3.2 1B and GSM8K, Dynamic K-Head reaches 54.44% exact-match accuracy at 12.03 average reasoning/output tokens, compared with 52.31% at 15.21 tokens for fixed-$K=6$ CoDI---a 2.12-point gain with 20.9% lower token cost. Across three out-of-distribution arithmetic datasets, it reduces cost by 19.3--25.7% while improving accuracy by 0.30--2.22 points. These results demonstrate that difficulty-supervised budget selection can improve both accuracy and efficiency, while stable budget selection remains a central challenge.

Abstract: On-policy distillation (OPD) strengthens language-model reasoning, yet whether students acquire new factual knowledge or compositional skill for multi-step reasoning remains unknown. We separate these capabilities using a controlled synthetic framework that measures the student's initial capabilities and independently controls the teacher's additional facts, compositional skill, or both. Across four models from three families, reverse-KL OPD reliably transfers compositional skill across unseen reasoning structures, but transfers minimal factual knowledge. Decoupling the distillation recipe reveals the source of this asymmetry: replacing reverse KL with forward KL restores factual transfer, whereas student rollouts specifically improve the execution of multi-step reasoning. Experiments on recent factual QA and competition mathematics show a similar asymmetry under reverse-KL OPD, yielding notable reasoning gains without factual memory expansion. Together, these results demonstrate that on-policy distillation does not expand a model's parametric knowledge, but instead teaches it to organize and compose the knowledge it already possesses.

Abstract: Reasoning models generate long chains of thought before they answer, yet it is debated whether the content of these chains does real computational work or is largely decorative. We study this question by viewing a transformer as a shallow circuit. One forward pass through a fixed number of layers has constant depth, so any procedure that runs the model a constant number of times is a shallow circuit. We call the best accuracy that a shallow circuit can reach on a task the ceiling of the task, and a task is serial if its ceiling lies below one. We prove three results on serial tasks that hold for every transformer, no matter how it was trained. Necessity: replacing the chain by anything that does not depend on its content, such as filler tokens or a restatement of the question, drives the accuracy down to the ceiling, and on a maximally serial task down to chance. Depth: no shallow computation can write the chain of a model whose accuracy exceeds the ceiling, not even approximately. Locality: the answer is one shallow pass away from the finished chain, so all of the serial reasoning happens in the chain. On word problems of finite groups, whose ceilings are known, small transformers trained from scratch, with or without reinforcement learning, attain the predicted numbers: chain-trained models solve every input length and fall to chance when the chain is erased, chainless models collapse to the ceiling as the input length grows, and open-weight reasoning models given the same problem in words return to the baseline without their chain. On MATH-500 and AIME, erasing the chain costs open reasoning models 0.52 to 0.82 accuracy, a sentence shuffle is harmless, and a token shuffle is as harmful as erasing; the same holds for checkpoints trained by GRPO with a correct or a random reward. The ceiling of a task therefore answers when a transformer can succeed without its chain of thought.