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.
☆32Updated 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:
- Large language models (LLMs) made easy, EasyLM is a one stop solution for pre-training, finetuning, evaluating and serving LLMs in JAX/Fl…☆76Updated last year
- Directional Preference Alignment☆58Updated last year
- ☆106Updated last year
- NeurIPS 2024 tutorial on LLM Inference☆47Updated 11 months ago
- Self-Supervised Alignment with Mutual Information☆21Updated last year
- Advantage Leftover Lunch Reinforcement Learning (A-LoL RL): Improving Language Models with Advantage-based Offline Policy Gradients☆26Updated last year
- This is code for most of the experiments in the paper Understanding the Effects of RLHF on LLM Generalisation and Diversity☆47Updated last year
- Code and Configs for Asynchronous RLHF: Faster and More Efficient RL for Language Models☆67Updated 7 months ago
- Reinforcing General Reasoning without Verifiers☆92Updated 5 months ago
- ICML 2024 - Official Repository for EXO: Towards Efficient Exact Optimization of Language Model Alignment☆57Updated last year
- The repository contains code for Adaptive Data Optimization☆28Updated 11 months ago
- ☆158Updated last year
- CodeUltraFeedback: aligning large language models to coding preferences (TOSEM 2025)☆72Updated last year
- Dateset Reset Policy Optimization☆31Updated last year
- Code for "Reasoning to Learn from Latent Thoughts"☆122Updated 8 months ago
- RL algorithm: Advantage induced policy alignment☆66Updated 2 years ago
- Official implementation of Bootstrapping Language Models via DPO Implicit Rewards☆44Updated 7 months ago
- official implementation of ICLR'2025 paper: Rethinking Bradley-Terry Models in Preference-based Reward Modeling: Foundations, Theory, and…☆69Updated 8 months ago
- ☆100Updated last year
- ☆33Updated 10 months ago
- Official repository of paper "RNNs Are Not Transformers (Yet): The Key Bottleneck on In-context Retrieval"☆27Updated last year
- Verlog: A Multi-turn RL framework for LLM agents☆66Updated 2 weeks ago
- Using FlexAttention to compute attention with different masking patterns☆47Updated last year
- [ICLR 2025] "Training LMs on Synthetic Edit Sequences Improves Code Synthesis" (Piterbarg, Pinto, Fergus)☆19Updated 9 months ago
- Code for Paper (Policy Optimization in RLHF: The Impact of Out-of-preference Data)☆28Updated last year
- Exploration of automated dataset selection approaches at large scales.☆50Updated 9 months ago
- Domain-specific preference (DSP) data and customized RM fine-tuning.☆25Updated last year
- ☆109Updated last year
- Replicating O1 inference-time scaling laws☆90Updated last year
- Code for "Echo Chamber: RL Post-training Amplifies Behaviors Learned in Pretraining"☆26Updated last month