RepairAgent autonomously repairs 164 bugs on Defects4J including 39 not fixed by prior techniques by treating an LLM as an agent that invokes tools via a finite state machine and dynamic prompts.
hub Canonical reference
PAL: Program-aided Language Models
Canonical reference. 83% of citing Pith papers cite this work as background.
34
Pith papers citing it
Background
83% of classified citations
abstract
Large language models (LLMs) have recently demonstrated an impressive ability to perform arithmetic and symbolic reasoning tasks, when provided with a few examples at test time ("few-shot prompting"). Much of this success can be attributed to prompting methods such as "chain-of-thought'', which employ LLMs for both understanding the problem description by decomposing it into steps, as well as solving each step of the problem. While LLMs seem to be adept at this sort of step-by-step decomposition, LLMs often make logical and arithmetic mistakes in the solution part, even when the problem is decomposed correctly. In this paper, we present Program-Aided Language models (PAL): a novel approach that uses the LLM to read natural language problems and generate programs as the intermediate reasoning steps, but offloads the solution step to a runtime such as a Python interpreter. With PAL, decomposing the natural language problem into runnable steps remains the only learning task for the LLM, while solving is delegated to the interpreter. We demonstrate this synergy between a neural LLM and a symbolic interpreter across 13 mathematical, symbolic, and algorithmic reasoning tasks from BIG-Bench Hard and other benchmarks. In all these natural language reasoning tasks, generating code using an LLM and reasoning using a Python interpreter leads to more accurate results than much larger models. For example, PAL using Codex achieves state-of-the-art few-shot accuracy on the GSM8K benchmark of math word problems, surpassing PaLM-540B which uses chain-of-thought by absolute 15% top-1. Our code and data are publicly available at http://reasonwithpal.com/ .
hub tools
citation-role summary
background 11
method 1
citation-polarity summary
representative citing papers
Introduces CaTS-Bench with human gold-standard captions and a synthetic generation pipeline to evaluate vision-language models on time series captioning and numeric reasoning.
ViperGPT generates executable Python code to compose pre-trained vision-and-language modules into programs that answer visual queries, reaching state-of-the-art results with no additional training.
ThinC trains small models to reason primarily in code rather than natural language, outperforming tool-integrated baselines and even larger models on competition math benchmarks.
Homogeneous multi-agent debate introduces sycophantic conformity, contextual fragility, and consensus collapse, leading to equal or lower accuracy than isolated self-correction at 2.1-3.4x higher token cost on GSM-Hard and MMLU-Hard.
LLMs match or exceed state-of-the-art traditional methods for stabilizing numerical expressions in scientific software, succeeding on 97.9% of expressions where baselines fail to improve accuracy, but struggle with control flow and high-precision literals.
LLM+P lets LLMs solve planning problems optimally by converting them to PDDL for classical planners and back to natural language.
PoT prompting improves numerical reasoning by having language models write programs executed by a computer instead of performing calculations in natural language chains of thought, with an average 12% gain over CoT.
Relabeling an identical erroneous claim from the model's own thought role to an external chat role increases explicit correction rates by 23-93 percentage points across 13 model-domain cells, indicating a chat-template artifact rather than a cognitive deficit.
CodeThinker improves LLM code reasoning via consistency-based RL with stepwise training data, dynamic beam sampling, and consistency rewards, reaching SOTA on benchmarks with 4.3% gains on Qwen2.5-Coder-7B.
Continued pretraining of Code Llama on Proof-Pile-2 yields Llemma, an open math-specialized LLM that beats known open base models on MATH and supports tool use plus formal proving out of the box.
ToRA trains language models on interactive tool-use trajectories with imitation learning and output shaping to integrate reasoning and external tools, yielding 13-19% gains on math datasets and new highs like 44.6% on MATH for a 7B model.
ReWOO decouples reasoning from tool observations in augmented language models, delivering 5x token efficiency and 4% higher accuracy on multi-step reasoning benchmarks like HotpotQA.
ART automatically generates multi-step reasoning programs with tool integration for LLMs, yielding substantial gains over few-shot and auto-CoT prompting on BigBench and MMLU while matching hand-crafted CoT on most tasks.
A metacognitive harness uses LLMs' pre- and post-solution self-monitoring signals to control test-time reasoning, raising pooled accuracy from 48.3% to 56.9% on text, code, and multimodal benchmarks.
GRIEF fuzzer finds 15 vulnerabilities including 2 CVEs in vLLM and SGLang by testing concurrent workloads for KV-cache isolation failures and cross-request interference.
QMP-Bench supplies a realistic test set for AI on quantum many-body problems while PhysVEC uses integrated verifiers to turn unreliable LLM generations into code that passes both syntax and physics checks, outperforming baselines.
ReTool uses outcome-driven RL to train 32B LLMs to dynamically use code tools during reasoning, reaching 72.5% accuracy on AIME and surpassing o1-preview.
An adaptive compute-optimal strategy for scaling LLM test-time compute achieves over 4x efficiency gains versus best-of-N and lets smaller models outperform 14x larger ones on some problems.
Gorilla is a fine-tuned LLM that surpasses GPT-4 in accurate API call generation and uses retrieval to handle documentation updates.
Self-Debugging teaches LLMs to identify and fix their own code errors through rubber-duck-style natural language explanations and execution feedback, delivering 2-12% gains over baselines on Spider, TransCoder, and MBPP.
HuggingGPT is an agent system where ChatGPT plans and orchestrates calls to Hugging Face models to solve complex multi-modal AI tasks.
MM-REACT uses textual prompts to let ChatGPT collaborate with external vision experts for zero-shot multimodal reasoning and action on advanced visual tasks.
RL post-training lifts answer correctness on FHIR-AgentBench from 50% (o4-mini) to 77% with a cheaper Qwen3-8B CodeAct agent.
citing papers explorer
-
RepairAgent: An Autonomous, LLM-Based Agent for Program Repair
RepairAgent autonomously repairs 164 bugs on Defects4J including 39 not fixed by prior techniques by treating an LLM as an agent that invokes tools via a finite state machine and dynamic prompts.
-
CaTS-Bench: Can Language Models Describe Time Series?
Introduces CaTS-Bench with human gold-standard captions and a synthetic generation pipeline to evaluate vision-language models on time series captioning and numeric reasoning.
-
ViperGPT: Visual Inference via Python Execution for Reasoning
ViperGPT generates executable Python code to compose pre-trained vision-and-language modules into programs that answer visual queries, reaching state-of-the-art results with no additional training.
-
Teaching Language Models to Think in Code
ThinC trains small models to reason primarily in code rather than natural language, outperforming tool-integrated baselines and even larger models on competition math benchmarks.
-
The Cost of Consensus: Isolated Self-Correction Prevails Over Unguided Homogeneous Multi-Agent Debate
Homogeneous multi-agent debate introduces sycophantic conformity, contextual fragility, and consensus collapse, leading to equal or lower accuracy than isolated self-correction at 2.1-3.4x higher token cost on GSM-Hard and MMLU-Hard.
-
Assessing Large Language Models for Stabilizing Numerical Expressions in Scientific Software
LLMs match or exceed state-of-the-art traditional methods for stabilizing numerical expressions in scientific software, succeeding on 97.9% of expressions where baselines fail to improve accuracy, but struggle with control flow and high-precision literals.
-
LLM+P: Empowering Large Language Models with Optimal Planning Proficiency
LLM+P lets LLMs solve planning problems optimally by converting them to PDDL for classical planners and back to natural language.
-
Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks
PoT prompting improves numerical reasoning by having language models write programs executed by a computer instead of performing calculations in natural language chains of thought, with an average 12% gain over CoT.
-
The Self-Correction Illusion: LLMs Correct Others but Not Themselves
Relabeling an identical erroneous claim from the model's own thought role to an external chat role increases explicit correction rates by 23-93 percentage points across 13 model-domain cells, indicating a chat-template artifact rather than a cognitive deficit.
-
Enhancing the Code Reasoning Capabilities of LLMs via Consistency-based Reinforcement Learning
CodeThinker improves LLM code reasoning via consistency-based RL with stepwise training data, dynamic beam sampling, and consistency rewards, reaching SOTA on benchmarks with 4.3% gains on Qwen2.5-Coder-7B.
-
Llemma: An Open Language Model For Mathematics
Continued pretraining of Code Llama on Proof-Pile-2 yields Llemma, an open math-specialized LLM that beats known open base models on MATH and supports tool use plus formal proving out of the box.
-
ToRA: A Tool-Integrated Reasoning Agent for Mathematical Problem Solving
ToRA trains language models on interactive tool-use trajectories with imitation learning and output shaping to integrate reasoning and external tools, yielding 13-19% gains on math datasets and new highs like 44.6% on MATH for a 7B model.
-
ReWOO: Decoupling Reasoning from Observations for Efficient Augmented Language Models
ReWOO decouples reasoning from tool observations in augmented language models, delivering 5x token efficiency and 4% higher accuracy on multi-step reasoning benchmarks like HotpotQA.
-
ART: Automatic multi-step reasoning and tool-use for large language models
ART automatically generates multi-step reasoning programs with tool integration for LLMs, yielding substantial gains over few-shot and auto-CoT prompting on BigBench and MMLU while matching hand-crafted CoT on most tasks.
-
LLMs Know When They Know, but Do Not Act on It: A Metacognitive Harness for Test-time Scaling
A metacognitive harness uses LLMs' pre- and post-solution self-monitoring signals to control test-time reasoning, raising pooled accuracy from 48.3% to 56.9% on text, code, and multimodal benchmarks.
-
Continuous Discovery of Vulnerabilities in LLM Serving Systems with Fuzzing
GRIEF fuzzer finds 15 vulnerabilities including 2 CVEs in vLLM and SGLang by testing concurrent workloads for KV-cache isolation failures and cross-request interference.
-
Towards Verifiable and Self-Correcting AI Physicists for Quantum Many-Body Simulations
QMP-Bench supplies a realistic test set for AI on quantum many-body problems while PhysVEC uses integrated verifiers to turn unreliable LLM generations into code that passes both syntax and physics checks, outperforming baselines.
-
ReTool: Reinforcement Learning for Strategic Tool Use in LLMs
ReTool uses outcome-driven RL to train 32B LLMs to dynamically use code tools during reasoning, reaching 72.5% accuracy on AIME and surpassing o1-preview.
-
Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters
An adaptive compute-optimal strategy for scaling LLM test-time compute achieves over 4x efficiency gains versus best-of-N and lets smaller models outperform 14x larger ones on some problems.
-
Gorilla: Large Language Model Connected with Massive APIs
Gorilla is a fine-tuned LLM that surpasses GPT-4 in accurate API call generation and uses retrieval to handle documentation updates.
-
Teaching Large Language Models to Self-Debug
Self-Debugging teaches LLMs to identify and fix their own code errors through rubber-duck-style natural language explanations and execution feedback, delivering 2-12% gains over baselines on Spider, TransCoder, and MBPP.
-
HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face
HuggingGPT is an agent system where ChatGPT plans and orchestrates calls to Hugging Face models to solve complex multi-modal AI tasks.
-
MM-REACT: Prompting ChatGPT for Multimodal Reasoning and Action
MM-REACT uses textual prompts to let ChatGPT collaborate with external vision experts for zero-shot multimodal reasoning and action on advanced visual tasks.
-
Reinforcement Learning for Tool-Calling Agents in Fast Healthcare Interoperability Resources (FHIR)
RL post-training lifts answer correctness on FHIR-AgentBench from 50% (o4-mini) to 77% with a cheaper Qwen3-8B CodeAct agent.
-
BRIDGE: Building Representations In Domain Guided Program Synthesis
BRIDGE improves Lean executable correctness up to 1.5x and sample efficiency roughly 2x over direct prompting by using domain-guided intermediate representations across code, specs, and proofs.
-
AdaSwitch: Adaptive Switching between Small and Large Agents for Effective Cloud-Local Collaborative Learning
AdaSwitch improves small local LLM performance on reasoning tasks by adaptively switching to a large cloud LLM upon detected errors, sometimes matching cloud results with far less overhead.
-
LLMs with in-context learning for Algorithmic Theoretical Physics
Frontier LLMs with in-context learning and CAS integration solve most algorithmic tasks in theoretical physics when supplied with worked examples.
-
The Cartesian Cut in Agentic AI
LLM agents use a Cartesian split between learned prediction and engineered control, enabling modularity but creating sensitivity and bottlenecks unlike integrated biological systems.
-
StarCoder: may the source be with you!
StarCoderBase matches or beats OpenAI's code-cushman-001 on multi-language code benchmarks; the Python-fine-tuned StarCoder reaches 40% pass@1 on HumanEval while retaining other-language performance.
-
Self-Refine: Iterative Refinement with Self-Feedback
Self-Refine boosts LLM outputs by ~20% on average across seven tasks by having the same model iteratively generate, critique, and refine its own responses.
-
Step-Video-T2V Technical Report: The Practice, Challenges, and Future of Video Foundation Model
Step-Video-T2V describes a 30B-parameter text-to-video model with custom Video-VAE, 3D DiT, flow matching, and Video-DPO that claims state-of-the-art results on a new internal benchmark.
-
SciFi: A Safe, Lightweight, User-Friendly, and Fully Autonomous Agentic AI Workflow for Scientific Applications
SciFi is a safe, lightweight agentic AI framework that automates structured scientific tasks with minimal human intervention via isolated environments and layered self-assessing agents.
-
A Survey on Multimodal Large Language Models
This survey organizes the architectures, training strategies, data, evaluation methods, extensions, and challenges of Multimodal Large Language Models.
- Correct Is Not Enough: Training Reasoning Planners with Executor-Grounded Rewards