SOTAVerified

Math Word Problem Solving

A math word problem is a mathematical exercise (such as in a textbook, worksheet, or exam) where significant background information on the problem is presented in ordinary language rather than in mathematical notation. As most word problems involve a narrative of some sort, they are sometimes referred to as story problems and may vary in the amount of technical language used.

Papers

Recently AddedMost HypedMost ActiveNeeds VerificationMost Verified

Showing 1–50 of 107 papers

TitleStatusHype
Qwen2 Technical ReportCode13
DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language ModelsCode9
Llama 2: Open Foundation and Fine-Tuned Chat ModelsCode8
LLaMA: Open and Efficient Foundation Language ModelsCode7
Mistral 7BCode6
Sparks of Artificial General Intelligence: Early experiments with GPT-4Code6
WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-InstructCode5
OpenMathInstruct-2: Accelerating AI for Math with Massive Open-Source Instruction DataCode4
OpenMathInstruct-1: A 1.8 Million Math Instruction Tuning DatasetCode4
Mixtral of ExpertsCode4
Let's Verify Step by StepCode4
Galactica: A Large Language Model for ScienceCode4
Step-DPO: Step-wise Preference Optimization for Long-chain Reasoning of LLMsCode3
AlphaMath Almost Zero: Process Supervision without ProcessCode3
Gemini 1.5: Unlocking multimodal understanding across millions of tokens of contextCode3
ToRA: A Tool-Integrated Reasoning Agent for Mathematical Problem SolvingCode3
PAL: Program-aided Language ModelsCode3
DART-Math: Difficulty-Aware Rejection Tuning for Mathematical Problem-SolvingCode2
MACM: Utilizing a Multi-Agent System for Condition Mining in Solving Complex Mathematical ProblemsCode2
GSM-Plus: A Comprehensive Benchmark for Evaluating the Robustness of LLMs as Mathematical Problem SolversCode2
An Empirical Study of Data Ability Boundary in LLMs' Math ReasoningCode2
Parameter-Efficient Sparsity Crafting from Dense to Mixture-of-Experts for Instruction Tuning on General TasksCode2
An Expression Tree Decoding Strategy for Mathematical Equation GenerationCode2
MuggleMath: Assessing the Impact of Query and Response Augmentation on Math ReasoningCode2
MathCoder: Seamless Code Integration in LLMs for Enhanced Mathematical ReasoningCode2
MetaMath: Bootstrap Your Own Mathematical Questions for Large Language ModelsCode2
Solving Challenging Math Word Problems Using GPT-4 Code Interpreter with Code-based Self-VerificationCode2
Cumulative Reasoning with Large Language ModelsCode2
Progressive-Hint Prompting Improves Reasoning in Large Language ModelsCode2
Multi-View Reasoning: Consistent Contrastive Learning for Math Word ProblemCode2
Solving Quantitative Reasoning Problems with Language ModelsCode2
Large Language Models are Zero-Shot ReasonersCode2
Measuring Mathematical Problem Solving With the MATH DatasetCode2
DeBERTa: Decoding-enhanced BERT with Disentangled AttentionCode2
Achieving >97% on GSM8K: Deeply Understanding the Problems Makes LLMs Better Solvers for Math Word ProblemsCode1
Toward Self-Improvement of LLMs via Imagination, Searching, and CriticizingCode1
Augmenting Math Word Problems via Iterative Question ComposingCode1
Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human AnnotationsCode1
FinanceMath: Knowledge-Intensive Math Reasoning in Finance DomainsCode1
Do Multilingual Language Models Think Better in English?Code1
MathChat: Converse to Tackle Challenging Math Problems with LLM AgentsCode1
Learning Multi-Step Reasoning by Solving Arithmetic TasksCode1
Automatic Model Selection with Large Language Models for ReasoningCode1
RetICL: Sequential Retrieval of In-Context Examples with Reinforcement LearningCode1
Automatic Generation of Socratic Subquestions for Teaching Math Word ProblemsCode1
ELASTIC: Numerical Reasoning with Adaptive Symbolic CompilerCode1
Learning to Reason Deductively: Math Word Problem Solving as Complex Relation ExtractionCode1
IconQA: A New Benchmark for Abstract Diagram Understanding and Visual Language ReasoningCode1
Recall and Learn: A Memory-augmented Solver for Math Word ProblemsCode1
MWPToolkit: An Open-Source Framework for Deep Learning-Based Math Word Problem SolversCode1
Show:102550
← PrevPage 1 of 3Next →
All datasetsMATHSVAMPMAWPSMath23KALG514ASDiv-AParaMAWPSDRAW-1KMathQASVAMP (1:N)GSM-PlusMATH minival

Benchmark Results

#ModelMetricClaimedVerifiedStatus
1Gemini 2.0 Flash ExperimentalAccuracy89.7—Unverified
2Qwen2.5-Math-72B-Instruct(TIR,Greedy)Accuracy88.1—Unverified
3GPT-4 Turbo (MACM, w/code, voting)Accuracy87.92—Unverified
4Qwen2.5-Math-72B-Instruct(COT,Greedy)Accuracy85.9—Unverified
5Qwen2.5-Math-7B-Instruct(TIR,Greedy)Accuracy85.2—Unverified
6GPT-4-code model (CSV, w/ code, SC, k=16)Accuracy84.3—Unverified
7Qwen2-Math-72B-Instruct(greedy)Accuracy84—Unverified
8Qwen2.5-Math-7B-Instruct(COT,Greedy)Accuracy83.6—Unverified
9Qwen2.5-Math-1.5B-Instruct(TIR,Greedy)Accuracy79.9—Unverified
10OpenMath2-Llama3.1-70B (majority@256)Accuracy79.6—Unverified
#ModelMetricClaimedVerifiedStatus
1GPT-4 DUPAccuracy94.2—Unverified
2GPT-4 (Teaching-Inspired)Execution Accuracy93.9—Unverified
3GPT-4 (Model Selection)Execution Accuracy93.7—Unverified
4Qwen2(CoT + Code Interpreter)Execution Accuracy92.3—Unverified
5GPT-4 (PHP)Execution Accuracy91.9—Unverified
6OpenMath-CodeLlama-70B (w/ code)Execution Accuracy87.8—Unverified
7MathCoder-L-70BExecution Accuracy84.9—Unverified
8PoT_Eng (self-consistency @ 5)Execution Accuracy83.7—Unverified
9CoT_Eng (self-consistency @ 5)Execution Accuracy82.5—Unverified
10MMOS-CODE-34B(0-shot)Execution Accuracy80.6—Unverified
#ModelMetricClaimedVerifiedStatus
1OpenMath-CodeLlama-70B (w/ code)Accuracy (%)95.7—Unverified
2MsAT-DeductReasonerAccuracy (%)94.3—Unverified
3ATHENA (roberta-large)Accuracy (%)93—Unverified
4Exp-TreeAccuracy (%)92.3—Unverified
5Multi-viewAccuracy (%)92.3—Unverified
6ATHENA (roberta-base)Accuracy (%)92.2—Unverified
7Roberta-DeductReasonerAccuracy (%)92—Unverified
8DeBERTa (PM + VM)Accuracy (%)91—Unverified
9EPTAccuracy (%)88.7—Unverified
10Graph2Tree with RoBERTaAccuracy (%)88.7—Unverified
#ModelMetricClaimedVerifiedStatus
1GPT-4 (Teaching-Inspired)Accuracy (5-fold)94.3—Unverified
2ATHENA (roberta-large)Accuracy (training-test)86.5—Unverified
3Multi-view* (ours)Accuracy (5-fold)85.2—Unverified
4ATHENA (roberta-base)Accuracy (training-test)84.4—Unverified
5Generate and RankAccuracy (5-fold)84.3—Unverified
6Exp-TreeAccuracy (5-fold)84.1—Unverified
7REAL2: Memory-augmented SolverAccuracy (5-fold)83.18—Unverified
8Roberta-DeductReasonerAccuracy (5-fold)83—Unverified
9MWP-BERTAccuracy (5-fold)82.4—Unverified
10Recall and LearnAccuracy (5-fold)80.8—Unverified