adp-v2

# ADP v2 This is an incremental ADP v2 dataset release generated from the `neulab/agent-data-protocol` pipeline. Each config corresponds to one source dataset. The initial upload contains only configs whose `raw`, `24k`, and `110k` stages are complete. Additional configs can be added later by committing their data files and updating this dataset card. ## Splits - `std`: standardized ADP/ATIF JSONL trajectories. - `sft_openhands_24k`: OpenHands SDK SFT records condensed with a 24k token target. - `sft_openhands_110k`: OpenHands SDK SFT records condensed with a 110k token target. ## Loading ```python from datasets import load_dataset std = load_dataset("neulab/adp-v2", "androidcontrol", split="std") sft24 = load_dataset("neulab/adp-v2", "androidcontrol", split="sft_openhands_24k") sft110 = load_dataset("neulab/adp-v2", "androidcontrol", split="sft_openhands_110k") ``` ## Initial Release Contents - Configs: 52 - Total `std` rows: 17676550 - Total `sft_openhands_24k` rows: 9640563 - Total `sft_openhands_110k` rows: 6353538 | Config | std rows | 24k SFT rows | 110k SFT rows | |---|---:|---:|---:| | `AlienKevin_SWE-ZERO-12M-trajectories` | 12290800 | 1001563 | 1000000 | | `CharlieDreemur_OpenManus-RL` | 48927 | 44647 | 44647 | | `SALT-NLP_SWE-chat` | 6122 | 40427 | 9497 | | `agenttuning_alfworld` | 336 | 336 | 336 | | `agenttuning_db` | 538 | 538 | 538 | | `agenttuning_kg` | 324 | 324 | 324 | | `agenttuning_mind2web` | 122 | 122 | 122 | | `agenttuning_os` | 195 | 195 | 195 | | `agenttuning_webshop` | 351 | 351 | 351 | | `allenai_Sera-4.6-Lite-T2` | 36083 | 120051 | 36083 | | `android_in_the_wild` | 5 | 5 | 5 | | `androidcontrol` | 3 | 3 | 3 | | `code_feedback` | 66383 | 66383 | 66383 | | `codeactinstruct` | 7139 | 7139 | 7139 | | `coderforge_preview` | 155144 | 709580 | 155830 | | `codescout` | 58871 | 62107 | 58871 | | `cognitivekernel_pro_sft` | 47271 | 47271 | 47271 | | `dolci_instruct_sft_tool_use` | 227579 | 227579 | 227603 | | `eto` | 6426 | 6425 | 6425 | | `finch_collection` | 217780 | 242234 | 217780 | | `gair_davinci_dev` | 74313 | 326010 | 46891 | | `go-browse-wa` | 8692 | 22164 | 9296 | | `hybrid-gym` | 4470 | 8076 | 4470 | | `jupyter-agent-dataset` | 51389 | 51429 | 51389 | | `kwai-klear_swe-smith-mini_swe_agent_plus-trajectories-66k` | 65994 | 36637 | 65992 | | `litecoder-terminal-sft` | 11255 | 32053 | 11522 | | `llava_plus` | 117039 | 117039 | 117039 | | `logicstar_swe-star` | 102761 | 389530 | 102762 | | `mind2web` | 1009 | 16125 | 9505 | | `mini-coder` | 84346 | 82406 | 66893 | | `miroverse_v0_1` | 147985 | 216207 | 148251 | | `nebius_SWE-agent-trajectories` | 80036 | 137154 | 80031 | | `nebius_SWE-rebench-openhands-trajectories` | 32161 | 181359 | 32125 | | `nemotron_terminal_corpus` | 44809 | 45106 | 44809 | | `nnetnav-live` | 48333 | 48333 | 48333 | | `nnetnav-wa` | 45277 | 45277 | 45277 | | `nvidia_SWE-Zero-openhands-trajectories` | 318115 | 956294 | 318168 | | `omniact` | 6789 | 6789 | 6789 | | `openhands` | 275 | 857 | 308 | | `openresearcher` | 97526 | 655476 | 100476 | | `openthoughts_agent_sft` | 94334 | 126566 | 94334 | | `openthoughts_tb_dev` | 36 | 36 | 36 | | `orca_agentinstruct` | 1046410 | 1046772 | 1046408 | | `scale_swe_distilled` | 71498 | 394399 | 71747 | | `screenagent` | 2000 | 2000 | 2000 | | `swe-gym_openhands_sampled_trajectories` | 6055 | 12186 | 6061 | | `swe-play-trajectories` | 704 | 2818 | 704 | | `swe-smith` | 32360 | 74658 | 32370 | | `synatra` | 99924 | 99924 | 99924 | | `toolmind` | 163180 | 163560 | 163180 | | `toucan_1_5m` | 1646546 | 1765409 | 1646515 | | `webarena_successful` | 530 | 634 | 530 |