paper-with-me

Papers

DISC: DISC: Dynamic Decomposition Improves LLM Inference Scaling

2025-02-23 · Jonathan Light, Wei Cheng, Benjamin Riviere, Wu Yue, Masafumi Oyamada, Mengdi Wang, Yisong Yue, Santiago Paternain, Haifeng Chen

Inference scaling methods for large language models often work by breaking problems into steps or groups of tokens, then sampling and selecting the best next steps. However, these steps and their sizes are usually fixed or manually designed based on domain knowledge. We introduce dynamic decomposition, a method that adaptively and automatically breaks down solution and reasoning traces into manageable steps during inference. By allocating compute more effectively - especially by subdividing difficult steps and prioritizing their sampling - dynamic decomposition significantly boosts inference efficiency. Experiments on benchmarks like APPS, MATH, and LiveCodeBench show that dynamic decomposition outperforms fixed strategies such as token-level, sentence-level, and single-step decompositions, reducing the pass@10 error rate by 5.0%, 6.7%, and 10.5% respectively. These results show the promise of dynamic decomposition for improving a broad range of inference scaling techniques.

📄 PDF Abstract BibTeX arXiv:2502.16706

Code (0)

등록된 구현이 없습니다.

Tasks

Computational EfficiencyMathSentence

Similar Papers 제목 키워드 기반

Stochastic Nonparametric Event-Tensor Decomposition

2018-12-01 · NeurIPS 2018 12 · Shandian Zhe, Yishuai Du

Tensor decompositions are fundamental tools for multiway data analysis. Existing approaches, however, ignore the valuable temporal information along with data, or simply discretize them into time steps so that important …

Tensor Decomposition

Dynamics-Aligned Shared Hypernetworks for Contextual RL under Discontinuous Shifts

2026-02-06 · Jan Benad, Pradeep Kr. Banerjee, Frank Röder, Nihat Ay 외 arxiv

Zero-shot generalization in contextual reinforcement learning remains a core challenge, particularly when the context is latent and must be inferred from data. A canonical failure mode arises when latent context disconti…

Zero-shot GeneralizationReinforcement Learning

Unsupervised Video Decomposition using Spatio-temporal Iterative Inference

2020-06-25 · Polina Zablotskaia, Edoardo A. Dominici, Leonid Sigal, Andreas M. Lehrmann

Unsupervised multi-object scene decomposition is a fast-emerging problem in representation learning. Despite significant progress in static scenes, such models are unable to leverage important dynamic cues present in vid…

ObjectRepresentation Learning

DCD: Decomposition-based Causal Discovery from Autocorrelated and Non-Stationary Temporal Data

2026-02-01 · Muhammad Hasan Ferdous, Md Osman Gani arxiv

Multivariate time series in domains such as finance, climate science, and healthcare often exhibit long-term trends, seasonal patterns, and short-term fluctuations, complicating causal inference under non-stationarity an…

Causal Inference

On Dynamic Programming Decompositions of Static Risk Measures in Markov Decision Processes

2023-04-24 · NeurIPS 2023 11 · Jia Lin Hau, Erick Delage, Mohammad Ghavamzadeh, Marek Petrik

Optimizing static risk-averse objectives in Markov decision processes is difficult because they do not admit standard dynamic programming equations common in Reinforcement Learning (RL) algorithms. Dynamic programming de…

Reinforcement Learning (RL)