Pith. sign in

REVIEW 17 cited by

Learning From Mistakes Makes LLM Better Reasoner

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 2310.20689 v4 pith:ZZVAMWKQ submitted 2023-10-31 cs.CL cs.AI

classification cs.CLcs.AI
keywords llmsdatalearninglemamistakereasoningmistakescapabilities
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Large language models (LLMs) recently exhibited remarkable reasoning capabilities on solving math problems. To further improve their reasoning capabilities, this work explores whether LLMs can LEarn from MistAkes (LEMA), akin to the human learning process. Consider a human student who failed to solve a math problem, he will learn from what mistake he has made and how to correct it. Mimicking this error-driven learning process, LEMA incorporates mistake-correction data pairs during fine-tuning LLMs. Specifically, we first collect inaccurate reasoning paths from various LLMs, and then employ GPT-4 as a ''corrector'' to identify the mistake step, explain the reason for the mistake, correct the mistake and generate the final answer. In addition, we apply a correction-centric evolution strategy that effectively expands the question set for generating correction data. Experiments across various LLMs and reasoning tasks show that LEMA effectively improves CoT-alone fine-tuning. Our further ablations shed light on the non-homogeneous effectiveness between CoT data and correction data. These results suggest a significant potential for LLMs to improve through learning from their mistakes. Our code, models and prompts are publicly available at https://github.com/microsoft/LEMA.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 17 Pith papers

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

  1. Toward Adaptive Reasoning in Large Language Models with Thought Rollback

    cs.AI 2024-12 conditional novelty 7.0 of 10

    Thought Rollback enables LLMs to revise prior reasoning steps through rollback and accumulated error analysis, improving solve rates on math and multi-task benchmarks while increasing token usage dramatically.

  2. PDD-RRG: Posterior Diagnostic Decision for Study-level Radiology Report Generation

    cs.CV 2026-08 conditional novelty 6.0 of 10

    A decision-level fusion method aggregates multiple draft radiology reports via Bayesian posterior scoring and validation-tuned thresholds, improving CheXbert F1 scores on MIMIC-CXR across three base models.

  3. SWE-Review: Closing the Loop on Issue Resolution with Agentic Code Review

    cs.SE 2026-07 conditional novelty 6.0 of 10

    An agentic code reviewer that explores repositories to judge and diagnose AI-generated pull requests improves resolve rates from 27.5% to 56.9% and outperforms single-turn review baselines.

  4. Failures Are the Stepping Stones to Success: Enhancing Few-Shot In-Context Learning by Leveraging Negative Samples

    cs.CL 2025-07 conditional novelty 6.0 of 10

    A retrieval method that uses semantically similar wrong answers as anchors to find corrective correct demonstrations improves few-shot in-context learning accuracy on seven reasoning datasets.

  5. Self-Correction Bench: Uncovering and Addressing the Self-Correction Blind Spot in Large Language Models

    cs.CL 2025-07 conditional novelty 6.0 of 10

    Non-reasoning LLMs fail to correct their own errors (64.5% blind spot) while correcting identical external errors, and appending 'Wait' cuts the gap by 89.3%.

  6. Self-Error-Instruct: Generalizing from Errors for LLMs Mathematical Reasoning

    cs.CL 2025-05 conditional novelty 6.0 of 10

    Self-Error-Instruct clusters a model's math errors into types, synthesizes targeted practice data per type, and selects the best samples for fine-tuning, improving held-out math test accuracy.

  7. T1: Advancing Language Model Reasoning through Reinforcement Learning and Inference Scaling

    cs.LG 2025-01 conditional novelty 6.0 of 10

    A training pipeline called T1 scales RL for LLM reasoning via diverse sampling and rewards, and claims that longer allowed generations directly improve math accuracy without verifiers.

  8. CDS: Knowledge Component-Driven Data Synthesis Guided by Cognitive Diagnosis Theory

    cs.AI 2025-01 conditional novelty 6.0 of 10

    A knowledge-component diagnostic pipeline, inspired by cognitive diagnosis theory, generates weakness-targeted synthetic data that improves small LLMs on math, code, and exam benchmarks by up to 13.1 percentage points.

  9. Repository Structure-Aware Training Makes SLMs Better Issue Resolver

    cs.SE 2024-12 conditional novelty 6.0 of 10

    Training 7B code models on automatically constructed repository-localization and code-edit data improves their SWE-Bench-verified resolution rate from about 1% to 7%.

  10. Error-driven Data-efficient Large Multimodal Model Tuning

    cs.CL 2024-12 conditional novelty 6.0 of 10

    An error-driven teacher-student pipeline extracts a student LMM's missing skills from validation mistakes and retrieves targeted samples from a task-agnostic dataset to fine-tune it.

  11. CRITICTOOL: Evaluating Self-Critique Capabilities of Large Language Models in Tool-Calling Error Scenarios

    cs.SE 2025-06 conditional novelty 5.0 of 10

    CRITICTOOL, a benchmark of 2,740 tool-calling error scenarios built from BFCL and T-Eval with GPT-4o-based error injection, finds most LLMs rarely recover from tool-use errors, with GPT-4o best at 69.01 overall and to...

  12. Exchange of Perspective Prompting Enhances Reasoning in Large Language Models

    cs.CL 2025-06 conditional novelty 5.0 of 10

    A two-branch prompting method that exchanges answers between an original math question and a paraphrased version improves accuracy on several math benchmarks, but the gain is not separated from the extra compute or ru...

  13. Step Back to Leap Forward: Self-Backtracking for Boosting Reasoning of Language Models

    cs.CL 2025-02 conditional novelty 5.0 of 10

    Training LLMs to append a <backtrack> token after erroneous reasoning steps and to use that signal at inference time improves Countdown accuracy by more than 40 percent over supervised fine-tuning.

  14. PSSD: Making Large Language Models Self-denial via Human Psyche Structure

    cs.AI 2025-02 conditional novelty 5.0 of 10

    A three-role (id, superego, ego) multi-agent prompting framework improves LLM reasoning accuracy by 1-4 points on multi-hop QA and math benchmarks.

  15. Reflection-Based Memory For Web navigation Agents

    cs.AI 2025-06 conditional novelty 4.0 of 10

    Reflection-Augmented Planning (ReAP) retrieves short self-reflections from past web navigation tasks and lifts held-out task success by 11 points on WebArena.

  16. A Comprehensive Survey on Imbalanced Data Learning

    cs.LG 2025-02 conditional novelty 3.0 of 10

    A structured survey and benchmark that groups imbalanced data learning methods into data re-balancing, feature representation, training strategy, and ensemble learning.

  17. 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