--- license: mit dataset_info: features: - name: task_type dtype: string - name: instance_index dtype: int64 - name: config dtype: string splits: - name: hill_default num_bytes: 639 num_examples: 2 - name: tree_default num_bytes: 53937 num_examples: 3 - name: maxsat_default num_bytes: 5820 num_examples: 4 - name: hill_hard num_bytes: 9781 num_examples: 7 - name: tree_hard num_bytes: 117947 num_examples: 7 - name: maxsat_hard num_bytes: 10608 num_examples: 7 - name: hill_full num_bytes: 20718 num_examples: 50 - name: tree_full num_bytes: 578119 num_examples: 50 - name: maxsat_full num_bytes: 128649 num_examples: 50 download_size: 499010 dataset_size: 926218 configs: - config_name: default data_files: - split: hill_default path: data/hill_default-* - split: tree_default path: data/tree_default-* - split: maxsat_default path: data/maxsat_default-* - split: hill_hard path: data/hill_hard-* - split: tree_hard path: data/tree_hard-* - split: maxsat_hard path: data/maxsat_hard-* - split: hill_full path: data/hill_full-* - split: tree_full path: data/tree_full-* - split: maxsat_full path: data/maxsat_full-* tags: - exploration - interactive - agent pretty_name: Explore-Exploit Benchmark size_categories: - n<1K --- # Explore-Exploit Bench — **Failing to Explore: Language Models on Interactive Tasks**
arXiv Website Github
![Visual Description of Interactive Tasks](https://raw.githubusercontent.com/mahdi-jfri/explore-exploit-bench/refs/heads/master/teaser.svg) We refer to [our github repo](https://github.com/mahdi-jfri/explore-exploit-bench) for more information on instances and how to use the dataset.