Logics-STEM: Empowering LLM Reasoning via Failure-Driven Post-Training and Document Knowledge Enhancement

2026-01-04

arXiv.org

AI Summary

This work addresses the limited complex reasoning capabilities of large language models in STEM domains by proposing a data-algorithm co-design paradigm. The authors construct a high-quality dataset comprising 10 million long-chain-of-thought samples and develop a five-stage data engine—encompassing annotation, deduplication, contamination removal, distillation, and stratified sampling—alongside a failure-driven post-training framework that effectively integrates open-source and synthetically generated data to optimize both supervised fine-tuning and reinforcement learning. Evaluated on an 8B-parameter model, the approach achieves an average performance gain of 4.68% over the strongest baseline on established STEM benchmarks. The project publicly releases both 8B and 32B models, together with 10M and 2.2M subsets of the curated dataset.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

We present Logics-STEM, a state-of-the-art reasoning model fine-tuned on Logics-STEM-SFT-Dataset, a high-quality and diverse dataset at 10M scale that represents one of the largest-scale open-source long chain-of-thought corpora. Logics-STEM targets reasoning tasks in the domains of Science, Technology, Engineering, and Mathematics (STEM), and exhibits exceptional performance on STEM-related benchmarks with an average improvement of 4.68% over the next-best model at 8B scale. We attribute the gains to our data-algorithm co-design engine, where they are jointly optimized to fit a gold-standard distribution behind reasoning. Data-wise, the Logics-STEM-SFT-Dataset is constructed from a meticulously designed data curation engine with 5 stages to ensure the quality, diversity, and scalability, including annotation, deduplication, decontamination, distillation, and stratified sampling. Algorithm-wise, our failure-driven post-training framework leverages targeted knowledge retrieval and data synthesis around model failure regions in the Supervised Fine-tuning (SFT) stage to effectively guide the second-stage SFT or the reinforcement learning (RL) for better fitting the target distribution. The superior empirical performance of Logics-STEM reveals the vast potential of combining large-scale open-source data with carefully designed synthetic data, underscoring the critical role of data-algorithm co-design in enhancing reasoning capabilities through post-training. We make both the Logics-STEM models (8B and 32B) and the Logics-STEM-SFT-Dataset (10M and downsampled 2.2M versions) publicly available to support future research in the open-source community.

Problem

• reasoning
• STEM
• large language models
• post-training
• chain-of-thought

Innovation

• failure-driven post-training
• data-algorithm co-design
• chain-of-thought reasoning
• synthetic data synthesis
• STEM reasoning