Reason-Wang/NAT
[NAACL 2025] The official implementation of paper "Learning From Failure: Integrating Negative Examples when Fine-tuning Large Language Models as Agents"
This project helps improve the reasoning capabilities of large language models (LLMs) used as AI agents. By integrating both successful and unsuccessful attempts at problem-solving, it teaches LLMs to avoid common pitfalls. The input is a collection of problem-solving attempts, and the output is a more robust LLM agent. This is for researchers and practitioners who are fine-tuning LLMs for complex tasks.
No commits in the last 6 months.
Use this if you are fine-tuning large language models to act as intelligent agents and want them to perform better on mathematical reasoning or complex question-answering tasks by learning from mistakes.
Not ideal if you are looking for a pre-trained general-purpose LLM without specific agentic fine-tuning needs or if your primary focus is on generative text rather than problem-solving.
Stars
28
Forks
—
Language
Python
License
—
Category
Last pushed
Mar 14, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/transformers/Reason-Wang/NAT"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
galilai-group/stable-pretraining
Reliable, minimal and scalable library for pretraining foundation and world models
CognitiveAISystems/MAPF-GPT
[AAAI-2025] This repository contains MAPF-GPT, a deep learning-based model for solving MAPF...
UKPLab/gpl
Powerful unsupervised domain adaptation method for dense retrieval. Requires only unlabeled...
larslorch/avici
Amortized Inference for Causal Structure Learning, NeurIPS 2022
svdrecbd/mhc-mlx
MLX + Metal implementation of mHC: Manifold-Constrained Hyper-Connections by DeepSeek-AI.