Pith. sign in

REVIEW 16 cited by

GSM-Plus: A Comprehensive Benchmark for Evaluating the Robustness of LLMs as Mathematical Problem Solvers

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 2402.19255 v2 pith:KBARRU3C submitted 2024-02-29 cs.CL

classification cs.CL
keywords llmsmathematicalreasoningmathachievedgsm-plusgsm8kmodels
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Large language models (LLMs) have achieved impressive performance across various mathematical reasoning benchmarks. However, there are increasing debates regarding whether these models truly understand and apply mathematical knowledge or merely rely on shortcuts for mathematical reasoning. One essential and frequently occurring evidence is that when the math questions are slightly changed, LLMs can behave incorrectly. This motivates us to evaluate the robustness of LLMs' math reasoning capability by testing a wide range of question variations. We introduce the adversarial grade school math (GSM-Plus) dataset, an extension of GSM8K augmented with various mathematical perturbations. Our experiments on 25 LLMs and 4 prompting techniques show that while LLMs exhibit different levels of math reasoning abilities, their performances are far from robust. In particular, even for problems that have been solved in GSM8K, LLMs can make mistakes when new statements are added or the question targets are altered. We also explore whether more robust performance can be achieved by composing existing prompting methods, in which we try an iterative method that generates and verifies each intermediate thought based on its reasoning goal and calculation result.

Discussion (0). Sign in to comment.

Forward citations

Cited by 16 Pith papers

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

  1. Why Do Multi-Agent LLM Systems Fail?

    cs.AI 2025-03 unverdicted novelty 8.0 of 10

    The authors create the first large-scale dataset and taxonomy of failure modes in multi-agent LLM systems to explain their limited performance gains.

  2. Search for Truth from Reasoning: A Dynamic Representation Editing Framework for Steering LLM Trajectories

    cs.AI 2026-06 unverdicted novelty 7.0 of 10

    DynaSteer dynamically steers LLM reasoning trajectories toward truth via pattern clustering, Fisher-LDA projection, and entropy-triggered representation edits, improving performance on MATH and generalizing to coding.

  3. Testing LLM Arithmetic Reasoning Generalization with Automatic Numeric-Remapping Attacks

    cs.CR 2026-06 unverdicted novelty 7.0 of 10

    An automatic numeric-remapping attack generator reveals 12-26 point accuracy drops on GSM8K for three LLMs while MAWPS and MultiArith stay near 98%.

  4. Remask, Don't Replace: Token-to-Mask Refinement in Diffusion Large Language Models

    cs.CL 2026-04 unverdicted novelty 7.0 of 10

    Token-to-Mask remasking improves self-correction in diffusion LLMs by resetting erroneous commitments to masks rather than overwriting them, yielding +13.33 points on AIME 2025 and +8.56 on CMATH.

  5. Do LLMs Overthink Basic Math Reasoning? Benchmarking the Accuracy-Efficiency Tradeoff in Language Models

    cs.CL 2025-07 conditional novelty 7.0 of 10

    Evaluations of 53 LLMs on 14 basic math tasks show reasoning models use ~18x more tokens with sometimes lower accuracy, non-monotonic gains from extended budgets, and sharp performance drops under token constraints.

  6. Are Arithmetic Heuristic Neurons Form-Invariant? A Mechanistic Analysis of Symbols, Text, and Code in LLMs

    cs.CL 2026-07 conditional novelty 6.0 of 10

    Llama-3 arithmetic is computed by a shared neuron set across symbolic, text, and code formats, and cross-format failures are activation-state differences, not distinct circuits.

  7. Implicit Reasoning Steering via Concept Chaining

    cs.CL 2026-07 conditional novelty 6.0 of 10

    Reinforcement-learning-optimized concept-chain paragraphs covertly steer language-model multiple-choice preferences after continued pretraining, with far lower detectability than direct paraphrases.

  8. Don't Commit Alone: Joint Token Commitment in Diffusion Large Language Models

    cs.CL 2026-07 conditional novelty 6.0 of 10

    Marker-gated re-application of a dLLM’s last layers lets a commit bundle coordinate before greedy write, improving accuracy on all six evaluated benchmarks under matched greedy inference.

  9. Search for Truth from Reasoning: A Dynamic Representation Editing Framework for Steering LLM Trajectories

    cs.AI 2026-06 unverdicted novelty 6.0 of 10

    DynaSteer is a dynamic representation editing framework that uses pattern clustering, Fisher-LDA, and lookahead entropy monitoring to steer LLM reasoning trajectories toward truth on MATH and coding tasks.

  10. Model soups need only one ingredient

    cs.LG 2026-02 conditional novelty 6.0 of 10

    A single checkpoint, edited by splitting each layer's update with SVD and reweighting the high- and low-energy parts, reaches soup-level OOD robustness without multi-model training.

  11. FLAMES: Improving LLM Math Reasoning via a Fine-Grained Analysis of the Data Synthesis Pipeline

    cs.LG 2025-08 conditional novelty 6.0 of 10

    A controlled study of synthetic math data generation yields a new dataset blend (FLAMES) that improves fine-tuned 7B model scores on MATH and OlympiadBench, with caveats on checkpoint selection.

  12. League of LLMs: A Benchmark-Free Paradigm for Mutual Evaluation of Large Language Models

    cs.AI 2025-07 unverdicted novelty 6.0 of 10

    League of LLMs organizes LLMs into a self-governed mutual evaluation league using dynamic, transparent, objective, and professional criteria to distinguish model capabilities with 70.7% top-k ranking stability.

  13. CoLD: Counterfactually-Guided Length Debiasing for Process Reward Models in Mathematical Reasoning

    cs.CL 2025-07 unverdicted novelty 6.0 of 10

    CoLD mitigates length bias in process reward models for mathematical reasoning via counterfactual guidance, length penalties, bias estimation, and joint training, improving step selection accuracy and conciseness on M...

  14. RTTC: Reward-Guided Collaborative Test-Time Compute

    cs.CL 2025-08 reject novelty 5.0 of 10

    RTTC uses a reward model to route each query to direct inference, retrieval-augmented generation, or test-time training, plus caching to cut redundant work.

  15. GSM-Plus-BN: A Perturbation-Based Benchmark for Bangla Mathematical Reasoning in Large Language Models

    cs.CL 2026-07 conditional novelty 4.0 of 10

    The paper releases GSM-Plus-BN, a human-verified Bengali translation of the GSM-Plus perturbed math benchmark, and reports accuracy baselines for six open LLMs under standard and CoT prompting.

  16. Mellum2 Technical Report

    cs.CL 2026-05 unverdicted novelty 3.0 of 10

    Mellum 2 is a 12B MoE model with 2.5B active parameters, trained on 10.6T tokens with MoE, GQA, SWA, and MTP, then post-trained into Instruct and Thinking variants, claimed competitive with 4B-14B models at 2.5B compute.

Pith tools