reward-bench
Reward model evaluator
A comprehensive benchmarking framework for evaluating the performance and safety of reward models in reinforcement learning.
RewardBench: the first evaluation tool for reward models.
459 stars
5 watching
54 forks
Language: Python
last commit: 11 months ago
Linked from 1 awesome list
preference-learningrlhf
Related projects:
| Repository | Description | Stars |
|---|---|---|
| allenai/olmo-eval | A framework for evaluating language models on NLP tasks | 326 |
 openai/simple-evals | Evaluates language models using standardized benchmarks and prompting techniques. | 2,059 |
 cloud-cv/evalai | A platform for comparing and evaluating AI and machine learning algorithms at scale | 1,779 |
 huggingface/evaluate | An evaluation framework for machine learning models and datasets, providing standardized metrics and tools for comparing model performance. | 2,063 |
 thudm/agentbench | A benchmark suite for evaluating the ability of large language models to operate as autonomous agents in various environments | 2,272 |
 chenllliang/mmevalpro | A benchmarking framework for evaluating Large Multimodal Models by providing rigorous metrics and an efficient evaluation pipeline. | 22 |
 ethicalml/xai | An eXplainability toolbox for machine learning that enables data analysis and model evaluation to mitigate biases and improve performance | 1,135 |
| allenai/tango | A software framework for organizing and running machine learning experiments with Python. | 533 |
 reworkd/bananalyzer | A tool to evaluate AI agents on web tasks by dynamically constructing and executing test suites against predefined example websites. | 274 |
 robustbench/robustbench | A standardized benchmark for measuring the robustness of machine learning models against adversarial attacks | 682 |
 carla-simulator/reinforcement-learning | An implementation of an actor-critic reinforcement learning algorithm in Python. | 245 |
 rll-research/url_benchmark | A benchmark suite for unsupervised reinforcement learning agents, providing pre-trained models and scripts for testing and fine-tuning agent performance. | 335 |
 maluuba/nlg-eval | A toolset for evaluating and comparing natural language generation models | 1,350 |
 open-evals/evals | A framework for evaluating OpenAI models and an open-source registry of benchmarks. | 19 |
 alco/benchfella | Tools for comparing and benchmarking small code snippets | 514 |