Pith. sign in

REVIEW 8 cited by

At Which Training Stage Does Code Data Help LLMs Reasoning?

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 2309.16298 v2 pith:UWHYQMYT submitted 2023-09-28 cs.CL

At Which Training Stage Does Code Data Help LLMs Reasoning?

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

Large Language Models (LLMs) have exhibited remarkable reasoning capabilities and become the foundation of language technologies. Inspired by the great success of code data in training LLMs, we naturally wonder at which training stage introducing code data can really help LLMs reasoning. To this end, this paper systematically explores the impact of code data on LLMs at different stages. Concretely, we introduce the code data at the pre-training stage, instruction-tuning stage, and both of them, respectively. Then, the reasoning capability of LLMs is comprehensively and fairly evaluated via six reasoning tasks in five domains. We critically analyze the experimental results and provide conclusions with insights. First, pre-training LLMs with the mixture of code and text can significantly enhance LLMs' general reasoning capability almost without negative transfer on other tasks. Besides, at the instruction-tuning stage, code data endows LLMs the task-specific reasoning capability. Moreover, the dynamic mixing strategy of code and text data assists LLMs to learn reasoning capability step-by-step during training. These insights deepen the understanding of LLMs regarding reasoning ability for their application, such as scientific question answering, legal support, etc. The source code and model parameters are released at the link:~\url{https://github.com/yingweima2022/CodeLLM}.

discussion (0)

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

Forward citations

Cited by 8 Pith papers

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

  1. Domain-Aware Scaling Laws Uncover Data Synergy

    cs.LG 2026-07 conditional novelty 7.0

    Domain-aware scaling laws with fitted γ and σ synergy terms recover stable code-math interactions from observational LLM mixtures and correctly predict mixture rankings in controlled small-scale trainings.

  2. Automatic Textbook Formalization

    cs.AI 2026-04 accept novelty 7.0

    Multi-agent AI system formalizes entire 500-page graduate algebraic combinatorics textbook into Lean, creating 130K lines of code in one week at human-expert cost.

  3. How Your Credentials Are Leaked by LLM Agent Skills: An Empirical Study

    cs.CR 2026-04 accept novelty 7.0

    Analysis of 17k LLM agent skills reveals 520 vulnerable ones with 1,708 leakage issues, primarily from debug output exposure, with a 10-pattern taxonomy and released dataset for future detection.

  4. How Your Credentials Are Leaked by LLM Agent Skills: An Empirical Study

    cs.CR 2026-04 conditional novelty 7.0

    Large-scale audit of SkillsMP agent skills finds 520 skills with 1,708 credential-leak issues, dominated by debug logging into the LLM context and hard-to-remediate forks.

  5. From Trainee to Trainer: LLM-Designed Training Environment for RL with Multi-Agent Reasoning

    cs.CL 2026-06 unverdicted novelty 6.0

    The LLM-as-Environment-Engineer framework lets the policy model redesign its own RL environments on the new MAPF-FrozenLake testbed, outperforming larger models and fixed baselines with Qwen3-4B.

  6. Attention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix It

    cs.CL 2026-06 conditional novelty 6.0

    CoT SFT disrupts long-range routing in hybrid models via changes to W_Q and W_K; QK-Restore restores pre-SFT projections to recover NIAH performance.

  7. What Really Improves Mathematical Reasoning: Structured Reasoning Signals Beyond Pure Code

    cs.AI 2026-05 unverdicted novelty 6.0

    Controlled experiments show structured reasoning traces and higher-density math-domain samples improve mathematical reasoning more than pure executable code, with internal routing patterns reflecting these data effects.

  8. Bridging Generation and Training: A Systematic Review of Quality Issues in LLMs for Code

    cs.SE 2026-05 accept novelty 6.0

    A review of 114 studies creates taxonomies for code and data quality issues, formalizes 18 propagation mechanisms from training data defects to LLM-generated code defects, and synthesizes detection and mitigation techniques.