Asap7772 / understanding-rlhfLinks
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.
☆29Updated last year
Alternatives and similar repositories for understanding-rlhf
Users that are interested in understanding-rlhf are comparing it to the libraries listed below
Sorting:
- Dateset Reset Policy Optimization☆30Updated last year
- Directional Preference Alignment☆56Updated 8 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
- Code for Paper (Policy Optimization in RLHF: The Impact of Out-of-preference Data)☆28Updated last year
- NeurIPS 2024 tutorial on LLM Inference☆45Updated 5 months ago
- Code and data used in the paper: "Training on Incorrect Synthetic Data via RL Scales LLM Math Reasoning Eight-Fold"☆30Updated 11 months ago
- ☆85Updated last year
- ☆49Updated 3 weeks ago
- official implementation of ICLR'2025 paper: Rethinking Bradley-Terry Models in Preference-based Reward Modeling: Foundations, Theory, and…☆58Updated 2 months ago
- Self-Supervised Alignment with Mutual Information☆18Updated last year
- Official implementation of Bootstrapping Language Models via DPO Implicit Rewards☆44Updated last month
- The repository contains code for Adaptive Data Optimization☆24Updated 5 months ago
- Code for "Echo Chamber: RL Post-training Amplifies Behaviors Learned in Pretraining"☆16Updated last month
- Large language models (LLMs) made easy, EasyLM is a one stop solution for pre-training, finetuning, evaluating and serving LLMs in JAX/Fl…☆75Updated 9 months ago
- ☆93Updated 11 months ago
- ☆13Updated last month
- The official implementation of Regularized Policy Gradient (RPG) (https://arxiv.org/abs/2505.17508)☆27Updated last week
- Reinforcing General Reasoning without Verifiers☆33Updated last week
- ☆140Updated 6 months ago
- Code for Paper (Preserving Diversity in Supervised Fine-tuning of Large Language Models)☆22Updated 3 weeks ago
- Domain-specific preference (DSP) data and customized RM fine-tuning.☆25Updated last year
- A Large-Scale, High-Quality Math Dataset for Reinforcement Learning in Language Models☆55Updated 3 months ago
- Code accompanying the paper "Noise Contrastive Alignment of Language Models with Explicit Rewards" (NeurIPS 2024)☆53Updated 6 months ago
- Online Adaptation of Language Models with a Memory of Amortized Contexts (NeurIPS 2024)☆63Updated 10 months ago
- Advantage Leftover Lunch Reinforcement Learning (A-LoL RL): Improving Language Models with Advantage-based Offline Policy Gradients☆26Updated 8 months ago
- Minimal RLHF implementation built on top of minGPT.☆28Updated 10 months ago
- Revisiting Mid-training in the Era of RL Scaling☆48Updated last month
- [ICLR 2025] "Training LMs on Synthetic Edit Sequences Improves Code Synthesis" (Piterbarg, Pinto, Fergus)☆19Updated 3 months ago
- ☆32Updated 4 months ago
- ☆97Updated 11 months ago