Pith. sign in

REVIEW 20 cited by

OpenMathInstruct-2: Accelerating AI for Math with Massive Open-Source Instruction Data

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

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 2410.01560 v2 pith:QGSTJI5J submitted 2024-10-02 cs.CL cs.AIcs.LG

Shubham Toshniwal , Wei Du , Ivan Moshkov , Branislav Kisacanin , Alexan Ayrapetyan , Igor Gitman This is my paper · ORCID
classification cs.CLcs.AIcs.LG
keywords datadatasetmathopenmathinstruct-2reasoningopen-sourcetextttaccess
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
abstract

Mathematical reasoning continues to be a critical challenge in large language model (LLM) development with significant interest. However, most of the cutting-edge progress in mathematical reasoning with LLMs has become \emph{closed-source} due to lack of access to training data. This lack of data access limits researchers from understanding the impact of different choices for synthesizing and utilizing the data. With the goal of creating a high-quality finetuning (SFT) dataset for math reasoning, we conduct careful ablation experiments on data synthesis using the recently released \texttt{Llama3.1} family of models. Our experiments show that: (a) solution format matters, with excessively verbose solutions proving detrimental to SFT performance, (b) data generated by a strong teacher outperforms equally-sized data generated by a weak student model, (c) SFT is robust to low-quality solutions, allowing for imprecise data filtering, and (d) question diversity is crucial for achieving data scaling gains. Based on these insights, we create the OpenMathInstruct-2 dataset, which consists of 14M question-solution pairs ($\approx$ 600K unique questions), making it nearly eight times larger than the previous largest open-source math reasoning dataset. Finetuning the \texttt{Llama-3.1-8B-Base} using OpenMathInstruct-2 outperforms \texttt{Llama3.1-8B-Instruct} on MATH by an absolute 15.9\% (51.9\% $\rightarrow$ 67.8\%). Finally, to accelerate the open-source efforts, we release the code, the finetuned models, and the OpenMathInstruct-2 dataset under a commercially permissive license.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 20 Pith papers

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

  1. Reference-Based Distillation Detection in LLMs

    cs.LG 2026-06 conditional novelty 7.5 of 10

    Reference-based membership inference recovers the true teacher of a distilled LLM with near-perfect accuracy when an earlier same-lineage checkpoint is available.

  2. Bridging Compute- and Data-Optimal Pretraining

    cs.LG 2026-07 conditional novelty 7.0 of 10

    Pretraining loss obeys a single law in which repeated or paraphrased tokens count as η(N, data-per-parameter, expansion-ratio) fresh tokens, with total effective data saturating as derived tokens grow.

  3. Making Open-Source Text LLM Watermarks Durable Against Merging

    cs.CL 2026-05 conditional novelty 7.0 of 10

    A new training method, Merge-Adversarial Training, makes watermarks embedded in open-source LLMs survive model merging, boosting post-merge detection by up to 51 percentage points.

  4. PoLoRA: A Preconditioned Orthogonalized LoRA Optimizer

    cs.LG 2026-07 conditional novelty 6.0 of 10

    A preconditioned, product-aware spectral optimizer for LoRA reaches Adam's final loss in 1.2–1.7x fewer steps on instruction-tuning, with ~3% per-step overhead.

  5. Multi-Turn On-Policy Distillation with Prefix Replay

    cs.LG 2026-07 conditional novelty 6.0 of 10

    ReOPD offline-distills multi-turn agentic LLMs via teacher-prefix replay plus step-decay sampling, matching online OPD accuracy at ≥4× speed with zero tool calls.

  6. Falcon-H1: A Family of Hybrid-Head Language Models Redefining Efficiency and Performance

    cs.CL 2025-07 conditional novelty 6.0 of 10

    Falcon-H1 reports competitive benchmark scores for a 0.5B to 34B family of parallel hybrid attention/Mamba-2 models, claiming 2x to 4x parameter efficiency versus dense transformers.

  7. TaskCraft: Automated Generation of Agentic Tasks

    cs.CL 2025-06 conditional novelty 6.0 of 10

    TaskCraft automatically generates verifiable, multi-tool agentic tasks with execution trajectories via atomic task construction and depth/width extension, and fine-tuning on the resulting 36,000-task dataset improves ...

  8. Establishing Trustworthy LLM Evaluation via Shortcut Neuron Analysis

    cs.CL 2025-06 conditional novelty 6.0 of 10

    Shortcut neuron patching suppresses benchmark-contamination shortcuts in LLMs and yields evaluation scores that strongly correlate with the external MixEval benchmark.

  9. Merge Hijacking: Backdoor Attacks to Model Merging of Large Language Models

    cs.CR 2025-05 conditional novelty 6.0 of 10

    Merge Hijacking is a backdoor attack that lets a malicious LLM checkpoint poison any model it is merged with while preserving normal behavior.

  10. ReCopilot: Reverse Engineering Copilot in Binary Analysis

    cs.CR 2025-05 conditional novelty 6.0 of 10

    ReCopilot is a 7B binary-analysis LLM that reports 13% higher average scores than existing tools on a new six-task benchmark, but the benchmark and artifacts are not released.

  11. CodeI/O: Condensing Reasoning Patterns via Code Input-Output Prediction

    cs.CL 2025-02 conditional novelty 6.0 of 10

    Training on 3.5M code input-output prediction tasks with natural-language chain-of-thought improves LLM performance on math, logic, symbolic, scientific, and commonsense reasoning benchmarks.

  12. Optimizing Temperature for Language Models with Multi-Sample Inference

    cs.LG 2025-02 conditional novelty 6.0 of 10

    Selecting the temperature at the entropy turning point of a language model's generated text yields near-optimal multi-sample inference accuracy without labeled validation data.

  13. Train Smarter, Not Longer: Memorization-Guided Data Reuse for Efficient LLM Training

    cs.LG 2026-07 conditional novelty 5.0 of 10

    Properly spaced multi-epoch reuse of high-quality data, guided by a measured memorization window, continues to improve LLM performance far beyond the common four-epoch heuristic.

  14. Simple Policy Gradients for Reasoning with Diffusion Language Models

    cs.LG 2025-10 reject novelty 5.0 of 10

    AGRPO makes GRPO-style policy gradients tractable for diffusion LLMs by Monte-Carlo sampling denoising timesteps, but the unbiasedness claim only holds for a step-level objective, not the token-level GRPO objective.

  15. Can Structured Templates Facilitate LLMs in Tackling Harder Tasks? : An Exploration of Scaling Laws by Difficulty

    cs.AI 2025-08 reject novelty 5.0 of 10

    Training on easy synthetic math data lowers accuracy on hard benchmarks, and the proposed SST framework, which teaches explicit procedural chains, aims to reverse that drop.

  16. EdgeLoRA: An Efficient Multi-Tenant LLM Serving System on Edge Devices

    cs.DC 2025-07 conditional novelty 5.0 of 10

    EdgeLoRA combines automatic adapter routing, LRU caching with a memory pool, and grouped LoRA batching to serve thousands of LoRA adapters on edge devices with up to 4x higher throughput than llama.cpp.

  17. MCTS-Refined CoT: High-Quality Fine-Tuning Data for LLM-Based Repository Issue Resolution

    cs.SE 2025-06 conditional novelty 5.0 of 10

    MCTS-REFINE uses tree search plus strict ground-truth matching to build chain-of-thought training data that lifts open-source LLM issue-resolution scores on SWE-bench.

  18. Improving Large Language Models with Concept-Aware Fine-Tuning

    cs.LG 2025-06 conditional novelty 5.0 of 10

    Adding lightweight multi-token auxiliary heads with a weighted future-token loss improves supervised fine-tuning of Llama-3-8B-Instruct across five diverse tasks.

  19. Logits-Based Finetuning

    cs.LG 2025-05 conditional novelty 5.0 of 10

    The proposed logits-based fine-tuning, which mixes teacher logits with ground truth labels, improves math reasoning accuracy of small LLaMA models over standard supervised fine-tuning, with a controlled GSM8K gain of ...

  20. A Survey on Large Language Models for Mathematical Reasoning

    cs.AI 2025-06 conditional novelty 1.0 of 10

    Recent advances in LLM mathematical reasoning are organized into comprehension and generation phases, covering methods from prompting to test-time scaling.

Pith tools