Asap7772 / understanding-rlhf
Learning from preferences is a common paradigm for fine-tuning language models. Yet, many algorithmic design decisions come into play. Our new work finds that approaches employing on-policy sampling or negative gradients outperform offline, maximum likelihood objectives.
☆28Updated last year
Alternatives and similar repositories for understanding-rlhf:
Users that are interested in understanding-rlhf are comparing it to the libraries listed below
- NeurIPS 2024 tutorial on LLM Inference☆43Updated 4 months ago
- ☆15Updated 5 months ago
- Self-Supervised Alignment with Mutual Information☆18Updated 11 months ago
- Directional Preference Alignment☆57Updated 7 months ago
- This is code for most of the experiments in the paper Understanding the Effects of RLHF on LLM Generalisation and Diversity☆43Updated last year
- [ICLR 2025] "Training LMs on Synthetic Edit Sequences Improves Code Synthesis" (Piterbarg, Pinto, Fergus)☆19Updated 2 months ago
- official implementation of ICLR'2025 paper: Rethinking Bradley-Terry Models in Preference-based Reward Modeling: Foundations, Theory, and…☆56Updated last month
- Dateset Reset Policy Optimization☆30Updated last year
- Advantage Leftover Lunch Reinforcement Learning (A-LoL RL): Improving Language Models with Advantage-based Offline Policy Gradients☆26Updated 7 months ago
- Minimal RLHF implementation built on top of minGPT.☆29Updated 10 months ago
- Code for the arXiv preprint "The Unreasonable Effectiveness of Easy Training Data"☆47Updated last year
- ☆31Updated 3 months ago
- A Large-Scale, High-Quality Math Dataset for Reinforcement Learning in Language Models☆51Updated 2 months ago
- Q-Probe: A Lightweight Approach to Reward Maximization for Language Models☆41Updated 10 months ago
- Code and Configs for Asynchronous RLHF: Faster and More Efficient RL for Language Models☆48Updated last week
- Code for Paper (Policy Optimization in RLHF: The Impact of Out-of-preference Data)☆28Updated last year
- Official repository of paper "RNNs Are Not Transformers (Yet): The Key Bottleneck on In-context Retrieval"☆27Updated last year
- Code for ICLR 2025 Paper "What is Wrong with Perplexity for Long-context Language Modeling?"☆58Updated last month
- ☆32Updated 4 months ago
- [ICLR 2025] Code for the paper "Beyond Autoregression: Discrete Diffusion for Complex Reasoning and Planning"☆52Updated 2 months ago
- Code accompanying the paper "Noise Contrastive Alignment of Language Models with Explicit Rewards" (NeurIPS 2024)☆51Updated 5 months ago
- ☆33Updated last month
- Efficient Scaling laws and collaborative pretraining.☆16Updated 3 months ago
- ☆84Updated last year
- ☆109Updated 3 months ago
- Dialogue Action Tokens: Steering Language Models in Goal-Directed Dialogue with a Multi-Turn Planner☆22Updated 10 months ago
- ☆91Updated 10 months ago
- Rewarded soups official implementation☆57Updated last year
- Repository for Skill Set Optimization☆12Updated 9 months ago
- ☆97Updated 10 months ago