Pith. sign in

REVIEW 26 cited by

Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2502.06703 v1 pith:MF7YEAFR submitted 2025-02-10 cs.CL

Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling

classification cs.CL
keywords modelspolicyapproachcomputationcompute-optimaldifficultyllmsmath-500
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Test-Time Scaling (TTS) is an important method for improving the performance of Large Language Models (LLMs) by using additional computation during the inference phase. However, current studies do not systematically analyze how policy models, Process Reward Models (PRMs), and problem difficulty influence TTS. This lack of analysis limits the understanding and practical use of TTS methods. In this paper, we focus on two core questions: (1) What is the optimal approach to scale test-time computation across different policy models, PRMs, and problem difficulty levels? (2) To what extent can extended computation improve the performance of LLMs on complex tasks, and can smaller language models outperform larger ones through this approach? Through comprehensive experiments on MATH-500 and challenging AIME24 tasks, we have the following observations: (1) The compute-optimal TTS strategy is highly dependent on the choice of policy model, PRM, and problem difficulty. (2) With our compute-optimal TTS strategy, extremely small policy models can outperform larger models. For example, a 1B LLM can exceed a 405B LLM on MATH-500. Moreover, on both MATH-500 and AIME24, a 0.5B LLM outperforms GPT-4o, a 3B LLM surpasses a 405B LLM, and a 7B LLM beats o1 and DeepSeek-R1, while with higher inference efficiency. These findings show the significance of adapting TTS strategies to the specific characteristics of each task and model and indicate that TTS is a promising approach for enhancing the reasoning abilities of LLMs.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 26 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Unsupervised Process Reward Models

    cs.LG 2026-05 unverdicted novelty 7.0

    Unsupervised PRMs derived from LLM probabilities achieve up to 15% better error detection than LLM judges and match supervised PRMs in verification and RL tasks.

  2. CA-SQL: Complexity-Aware Inference Time Reasoning for Text-to-SQL via Exploration and Compute Budget Allocation

    cs.CL 2026-05 unverdicted novelty 7.0

    CA-SQL achieves 51.72% execution accuracy on the challenging tier of the BIRD benchmark using GPT-4o-mini by scaling exploration breadth according to estimated task difficulty, evolutionary prompt seeding, and candida...

  3. Rewarding the Scientific Process: Process-Level Reward Modeling for Agentic Data Analysis

    cs.CL 2026-04 unverdicted novelty 7.0

    DataPRM is an environment-aware generative process reward model that improves LLM data analysis agents by 7-11% on benchmarks via active verification and reflection-aware ternary rewards.

  4. Rewarding the Scientific Process: Process-Level Reward Modeling for Agentic Data Analysis

    cs.CL 2026-04 unverdicted novelty 7.0

    DataPRM is a new process reward model for data analysis agents that detects silent errors via environment interaction and ternary rewards, yielding 7-11% gains on benchmarks and further RL improvements.

  5. TokenGS: Decoupling 3D Gaussian Prediction from Pixels with Learnable Tokens

    cs.CV 2026-04 unverdicted novelty 7.0

    TokenGS uses learnable Gaussian tokens in an encoder-decoder architecture to regress 3D means directly, achieving SOTA feed-forward reconstruction on static and dynamic scenes with better robustness.

  6. LLMs as a Jury: Cross-Model Consensus Can Outperform Process Reward Models for LLM Reasoning

    cs.LG 2026-07 conditional novelty 6.0

    Cross-model answer agreement is a free Best-of-N verifier that matches trained process reward models and is predicted by a parameter-free law from three panel statistics.

  7. When More Sampling Hurts: The Modal Ceiling and Correlation Ceiling of Test-Time Scaling

    cs.LG 2026-06 unverdicted novelty 6.0

    Test-time sampling improves coverage but stalls at modal and correlation ceilings for answer selection, with the effective number of samples as the practical limit.

  8. On-Policy Self-Distillation with Sampled Demonstrations Reduces Output Diversity

    cs.LG 2026-06 unverdicted novelty 6.0

    On-policy self-distillation with sampled demonstrations reduces rollout diversity by amplifying existing probability gaps in the base model, unlike ideal RL which preserves ratios among correct outputs.

  9. Breaking the Reward Barrier: Accelerating Tree-of-Thought Reasoning via Speculative Exploration

    cs.LG 2026-05 unverdicted novelty 6.0

    SPEX accelerates Tree-of-Thought LLM reasoning 1.2-3x via speculative path selection, dynamic budget allocation across queries, and adaptive early termination, with up to 4.1x when combined with token speculative decoding.

  10. Breaking the Reward Barrier: Accelerating Tree-of-Thought Reasoning via Speculative Exploration

    cs.LG 2026-05 unverdicted novelty 6.0

    SPEX delivers 1.2-3x speedup on ToT algorithms via speculative path selection, dynamic budget allocation, and adaptive early termination, reaching up to 4.1x when combined with token-level speculative decoding.

  11. Stream-T1: Test-Time Scaling for Streaming Video Generation

    cs.CV 2026-05 unverdicted novelty 6.0

    Stream-T1 is a test-time scaling framework for streaming video generation using scaled noise propagation from history, reward pruning across short and long windows, and feedback-guided memory sinking to improve tempor...

  12. Distilling Long-CoT Reasoning through Collaborative Step-wise Multi-Teacher Decoding

    cs.AI 2026-05 unverdicted novelty 6.0

    CoRD uses collaborative multi-teacher step-wise decoding with perplexity-guided beam search to generate higher-quality Long-CoT data that lets smaller models reach near-teacher performance with less supervision.

  13. Adaptive Test-Time Compute Allocation for Reasoning LLMs via Constrained Policy Optimization

    cs.LG 2026-04 unverdicted novelty 6.0

    A Lagrangian-relaxation plus imitation-learning pipeline adaptively allocates test-time compute to LLMs, outperforming uniform baselines by up to 12.8% relative accuracy on MATH while staying within a fixed average budget.

  14. DeepStack: Scalable and Accurate Design Space Exploration for Distributed 3D-Stacked AI Accelerators

    cs.AR 2026-04 conditional novelty 6.0

    DeepStack introduces a fast performance model and hierarchical search method for co-optimizing 3D DRAM stacking, interconnects, and distributed scheduling in AI accelerators, delivering up to 9.5x throughput gains ove...

  15. Decoding the Critique Mechanism in Large Reasoning Models

    cs.LG 2026-03 unverdicted novelty 6.0

    By injecting arithmetic mistakes into CoT reasoning, the paper identifies a hidden critique ability in LRMs and extracts a steerable critique vector that enhances self-correction across model scales.

  16. Do Not Waste Your Rollouts: Recycling Search Experience for Efficient Test-Time Scaling

    cs.CL 2026-01 unverdicted novelty 6.0

    RSE distills search trajectories into an experience bank for positive and negative recycling, yielding efficiency gains over independent sampling on math reasoning benchmarks.

  17. When Importance Sampling Misallocates Credit: Asymmetric Ratios for Outcome-Supervised RL

    cs.CL 2025-10 unverdicted novelty 6.0

    The paper identifies that importance sampling ratios in outcome-supervised RL misallocate credit by creating unbalanced token updates, and introduces ASPO to correct the asymmetry for positive-advantage tokens.

  18. Fin-PRM: A Domain-Specialized Process Reward Model for Financial Reasoning in Large Language Models

    cs.CL 2025-08 unverdicted novelty 6.0

    Fin-PRM is a domain-specialized process reward model that supplies binary step-level and trajectory-level supervision signals for financial reasoning in LLMs and outperforms general PRMs on CFLUE and FinQA benchmarks.

  19. Stabilizing Knowledge, Promoting Reasoning: Dual-Token Constraints for RLVR

    cs.CL 2025-07 unverdicted novelty 6.0

    Archer introduces response-level entropy normalization and differentiated clipping/KL regularization in RLVR to encourage exploration on reasoning tokens while stabilizing knowledge tokens, yielding gains in pass@1 an...

  20. OpenVLThinker: Complex Vision-Language Reasoning via Iterative SFT-RL Cycles

    cs.CV 2025-03 conditional novelty 6.0

    Iterative SFT-RL cycles enable a 7B LVLM to develop sophisticated visual chain-of-thought reasoning and improve performance on math and general reasoning benchmarks.

  21. Test-Time Scaling for Small VLMs on Multilingual Visual MCQ

    cs.CL 2026-07 accept novelty 5.5

    On EXAMS-V, token budget and parseability dominate chain count, PRM-guided search, and selectors for small VLMs; the policy model itself yields the largest gain.

  22. Proxy Exploration and Reusable Guidance: A Modular LLM Post-Training Paradigm via Proxy-Guided Update Signals

    cs.LG 2026-07 conditional novelty 5.0

    Relative policy-improvement signals from a weak proxy model, after simple calibration, can be transferred to improve stronger primary LLMs without re-exploring on the primary.

  23. Data Selection Through Iterative Self-Filtering for Vision-Language Settings

    cs.CV 2026-06 unverdicted novelty 5.0

    An iterative bootstrapped self-filtering approach selects balanced clean and diverse subsets from noisy vision-language datasets to train improved CLIP models.

  24. When RL Fails after SFT: Rejuvenating Model Plasticity for Robust SFT-to-RL Handoff

    cs.LG 2026-06 unverdicted novelty 5.0

    Excessive SFT reduces LLM plasticity for RL; Rejuvenation restores it via base-anchored fusion and targeted neuron resets, yielding better RL performance and OOD generalization.

  25. Trust Region On-Policy Distillation

    cs.LG 2026-05 unverdicted novelty 5.0

    TrOPD stabilizes on-policy distillation for LLMs with trust-region learning, outlier estimation, and off-policy guidance, outperforming prior OPD methods on reasoning and code benchmarks.

  26. Explicit Reasoning Makes Better Judges: A Systematic Study on Accuracy, Efficiency, and Robustness

    cs.AI 2025-09 unverdicted novelty 5.0

    Thinking LLMs achieve ~10 percentage points higher accuracy than non-thinking ones on RewardBench with under 2x compute overhead, outperforming augmentation strategies that cost over 8x more while also showing better ...