Pith. sign in

REVIEW 8 cited by

To Code, or Not To Code? Exploring Impact of Code in Pre-training

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 2408.10914 v1 pith:UICX466V submitted 2024-08-20 cs.CL

classification cs.CL
keywords codepre-trainingtasksimpactdataacrossperformancework
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Including code in the pre-training data mixture, even for models not specifically designed for code, has become a common practice in LLMs pre-training. While there has been anecdotal consensus among practitioners that code data plays a vital role in general LLMs' performance, there is only limited work analyzing the precise impact of code on non-code tasks. In this work, we systematically investigate the impact of code data on general performance. We ask "what is the impact of code data used in pre-training on a large variety of downstream tasks beyond code generation". We conduct extensive ablations and evaluate across a broad range of natural language reasoning tasks, world knowledge tasks, code benchmarks, and LLM-as-a-judge win-rates for models with sizes ranging from 470M to 2.8B parameters. Across settings, we find a consistent results that code is a critical building block for generalization far beyond coding tasks and improvements to code quality have an outsized impact across all tasks. In particular, compared to text-only pre-training, the addition of code results in up to relative increase of 8.2% in natural language (NL) reasoning, 4.2% in world knowledge, 6.6% improvement in generative win-rates, and a 12x boost in code performance respectively. Our work suggests investments in code quality and preserving code during pre-training have positive impacts.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 8 Pith papers

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

  1. Can Transformers Really Do It All? On the Compatibility of Inductive Biases Across Tasks

    cs.LG 2026-07 conditional novelty 7.0 of 10

    Learned replacement non-linearities show transformers are rarely optimal for algorithmic tasks, with benefits that are task-specific, while language/code gains are smaller and more transferable.

  2. Domain-Aware Scaling Laws Uncover Data Synergy

    cs.LG 2026-07 conditional novelty 7.0 of 10

    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.

  3. Can Large Language Models Generalize Procedures Across Representations?

    cs.CL 2026-02 conditional novelty 6.0 of 10

    Post-training on graph or code versions of a planning task does not transfer to natural-language versions, but a symbolic-then-natural-language RL curriculum achieves strong transfer.

  4. The Common Pile v0.1: An 8TB Dataset of Public Domain and Openly Licensed Text

    cs.CL 2025-06 conditional novelty 6.0 of 10

    A new 8TB openly-licensed text corpus trains 7B LLMs that are competitive with Llama 1/2, showing that performant models need not depend on unlicensed web data.

  5. How Programming Concepts and Neurons Are Shared in Code Language Models

    cs.CL 2025-06 conditional novelty 6.0 of 10

    In Llama-based code models, programming languages are represented through an English-like intermediate token space, with language-specific neurons concentrated in bottom layers and exclusive PL neurons in top layers; ...

  6. Transformers Pretrained on Procedural Data Contain Modular Structures for Algorithmic Reasoning

    cs.LG 2025-05 conditional novelty 6.0 of 10

    Different procedural pretraining tasks create complementary, transferable structures in a transformer's attention and MLP weights, and structures from different tasks can be combined into one initialization.

  7. What I cannot execute, I do not understand: Training and Evaluating LLMs on Program Execution Traces

    cs.LG 2025-02 conditional novelty 6.0 of 10

    Training an 8B LLM on Python execution traces improves output prediction on CruxEval and MBPP to about 80%, with dynamic scratchpads extending accurate predictions to executions of over ten thousand steps.

  8. Procedural Pretraining: Warming Up Language Models with Abstract Data

    cs.CL 2026-01 conditional novelty 5.0 of 10

    A short warm-up on procedural data (brackets, sorting, sets) makes language models more accurate and more data-efficient on language, code, and informal math.

Pith tools