snowman0919/qwen38-executor-27b-checkpoints-v2
Updated
id stringlengths 30 30 | split stringclasses 1
value | record_type stringclasses 1
value | messages listlengths 8 94 | tools listlengths 1 1 | images listlengths 3 46 | media listlengths 0 0 | artifacts listlengths 0 0 | task_type stringclasses 1
value | language stringclasses 1
value | domain stringclasses 23
values | difficulty stringclasses 1
value | source_repo stringclasses 1
value | source_revision stringclasses 1
value | source_config stringclasses 1
value | source_split stringclasses 1
value | source_row_id stringlengths 36 140 | source_url stringclasses 1
value | source_license stringclasses 1
value | trajectory_id stringlengths 36 140 | conversation_id stringlengths 24 24 | reward float64 0.86 1.43 | verified bool 1
class | verification_type stringclasses 1
value | derived bool 1
class | derived_from listlengths 0 0 | transformation listlengths 3 3 | content_hash stringlengths 64 64 | normalized_hash stringlengths 64 64 | prompt_hash stringlengths 64 64 | provenance stringlengths 195 299 | metadata stringlengths 110 183 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
train-db47faa220a57d9fc72b53c4 | train | computer_use_trajectory | [{"role":"system","content":[{"type":"text","text":"You are a cross-platform computer-use agent. Use(...TRUNCATED) | ["{\"function\":{\"description\":\"Execute grounded PyAutoGUI code.\",\"name\":\"computer_use\",\"pa(...TRUNCATED) | [{"bytes":"iVBORw0KGgoAAAANSUhEUgAABKgAAAMCCAIAAADCn2l0AAEAAElEQVR4nOydd5xdRfn/n5nTbt/ee7Jpm95JIBBC7(...TRUNCATED) | [] | [] | desktop_gui | en | Web Tools & Internet Utilities | xlangai/AgentNet | d76ee50a63fad81cfdbe576416757d7c2091ed50 | raw | train | 20240923010833_927af299-f6dc-4686-8221-b7f29cbe90b8 | https://huggingface.co/datasets/xlangai/AgentNet/tree/d76ee50a63fad81cfdbe576416757d7c2091ed50 | mit | 20240923010833_927af299-f6dc-4686-8221-b7f29cbe90b8 | db47faa220a57d9fc72b53c4 | 1.428571 | true | human_annotation_metadata_ok_high_alignment_unambiguous_completed_all_steps_correct_nonredundant | false | [] | [
"canonicalized_without_sampling",
"quality_gated",
"deduplicated"
] | 694d8125e2397c6920dfd75b3ce510a1c63e44e073aa67671d2d7e389e43c8b9 | cd0bfd213b01b684e3351c477029c4aaa9c5d48a9e6bd1b70800e03880c93147 | 1f68bf405d9f00aaecc138833a482aa30e1596ee31cf8c3bfa4bddd8764bda0c | "{\"source_config\":\"raw\",\"source_repo\":\"xlangai/AgentNet\",\"source_revision\":\"d76ee50a63fad(...TRUNCATED) | "{\"alignment_score\":10,\"applications\":[\"google chrome\"],\"difficulty\":3,\"efficiency_score\":(...TRUNCATED) | |
train-fc0206f7bd2a4445d524e6eb | train | computer_use_trajectory | [{"role":"system","content":[{"type":"text","text":"You are a cross-platform computer-use agent. Use(...TRUNCATED) | ["{\"function\":{\"description\":\"Execute grounded PyAutoGUI code.\",\"name\":\"computer_use\",\"pa(...TRUNCATED) | [{"bytes":"iVBORw0KGgoAAAANSUhEUgAABLwAAAL2CAIAAADErhhdAAEAAElEQVR4nFz9Pa8l25IthkXE/MzMtdbeVXXu7e73W(...TRUNCATED) | [] | [] | desktop_gui | en | Education & Research | xlangai/AgentNet | d76ee50a63fad81cfdbe576416757d7c2091ed50 | raw | train | 20240927225425_99161885-a201-4bb2-8828-0158a4c0200d | https://huggingface.co/datasets/xlangai/AgentNet/tree/d76ee50a63fad81cfdbe576416757d7c2091ed50 | mit | 20240927225425_99161885-a201-4bb2-8828-0158a4c0200d | fc0206f7bd2a4445d524e6eb | 1.428571 | true | human_annotation_metadata_ok_high_alignment_unambiguous_completed_all_steps_correct_nonredundant | false | [] | [
"canonicalized_without_sampling",
"quality_gated",
"deduplicated"
] | 15ded30557d709b98b55275f5eeab56121b36055c98c065a1de80820955282f2 | ddf3747284c7a5fac26289cde6f659395e961a53ef31b826a16fb840263936b1 | 627146f3619bef7cf40fd955af250ceb4185b3c6a5c5ef97ddc0ed86b56cac38 | "{\"source_config\":\"raw\",\"source_repo\":\"xlangai/AgentNet\",\"source_revision\":\"d76ee50a63fad(...TRUNCATED) | "{\"alignment_score\":10,\"applications\":[\"zotero\"],\"difficulty\":3,\"efficiency_score\":10,\"sy(...TRUNCATED) | |
train-56bf401eace734918c4fb003 | train | computer_use_trajectory | [{"role":"system","content":[{"type":"text","text":"You are a cross-platform computer-use agent. Use(...TRUNCATED) | ["{\"function\":{\"description\":\"Execute grounded PyAutoGUI code.\",\"name\":\"computer_use\",\"pa(...TRUNCATED) | [{"bytes":"iVBORw0KGgoAAAANSUhEUgAABQAAAALQCAIAAABAH0oBAAEAAElEQVR4nOz9d5xcx3UgCp9TdVOn6ckBM5hBzgADm(...TRUNCATED) | [] | [] | desktop_gui | en | E-commerce & Travel | xlangai/AgentNet | d76ee50a63fad81cfdbe576416757d7c2091ed50 | raw | train | 20240924070634_e7c07fe2-0125-4670-a660-37109ccf55fa | https://huggingface.co/datasets/xlangai/AgentNet/tree/d76ee50a63fad81cfdbe576416757d7c2091ed50 | mit | 20240924070634_e7c07fe2-0125-4670-a660-37109ccf55fa | 56bf401eace734918c4fb003 | 1.428571 | true | human_annotation_metadata_ok_high_alignment_unambiguous_completed_all_steps_correct_nonredundant | false | [] | [
"canonicalized_without_sampling",
"quality_gated",
"deduplicated"
] | 745a50046720621710ca362d24b2bcd711013d16e27f28e6edace83c6ea12eaa | d390a559700b8f2e5d79c46379b210e87c5ab976ac91f3ff7dfbbe42d72d9c4f | 6b1fd0da76ad49e4839a55804d91f00e2feccf4aada58f47a8da3e59c81455e5 | "{\"source_config\":\"raw\",\"source_repo\":\"xlangai/AgentNet\",\"source_revision\":\"d76ee50a63fad(...TRUNCATED) | "{\"alignment_score\":10,\"applications\":[],\"difficulty\":3,\"efficiency_score\":10,\"system\":\"D(...TRUNCATED) | |
train-f343a84e98d4239feb6f8b09 | train | computer_use_trajectory | [{"role":"system","content":[{"type":"text","text":"You are a cross-platform computer-use agent. Use(...TRUNCATED) | ["{\"function\":{\"description\":\"Execute grounded PyAutoGUI code.\",\"name\":\"computer_use\",\"pa(...TRUNCATED) | [{"bytes":"iVBORw0KGgoAAAANSUhEUgAABQAAAALQCAIAAABAH0oBAAEAAElEQVR4nOx9d7xdVZX/WnvvU255Pb33HlIIgdClC(...TRUNCATED) | [] | [] | desktop_gui | en | Social Media & Communication | xlangai/AgentNet | d76ee50a63fad81cfdbe576416757d7c2091ed50 | raw | train | 20240927011310_f74dd9bb-a435-4400-92eb-d5e235a83e04 | https://huggingface.co/datasets/xlangai/AgentNet/tree/d76ee50a63fad81cfdbe576416757d7c2091ed50 | mit | 20240927011310_f74dd9bb-a435-4400-92eb-d5e235a83e04 | f343a84e98d4239feb6f8b09 | 1.428571 | true | human_annotation_metadata_ok_high_alignment_unambiguous_completed_all_steps_correct_nonredundant | false | [] | [
"canonicalized_without_sampling",
"quality_gated",
"deduplicated"
] | d887abd9721e4b6b8751a6dca0419f09ee127d897f42416c8a1f00b2281f68c7 | dffda2ed7fe7fc073849b6100ce0be5fe5691344401066b9e192aeeea0d4bfbb | a692578b173e57987d95662e0427059c3153fcb0f1fa882f84f87904b87de6ab | "{\"source_config\":\"raw\",\"source_repo\":\"xlangai/AgentNet\",\"source_revision\":\"d76ee50a63fad(...TRUNCATED) | "{\"alignment_score\":10,\"applications\":[\"discord\"],\"difficulty\":6,\"efficiency_score\":10,\"s(...TRUNCATED) | |
train-8f60c320bae01534643436e8 | train | computer_use_trajectory | [{"role":"system","content":[{"type":"text","text":"You are a cross-platform computer-use agent. Use(...TRUNCATED) | ["{\"function\":{\"description\":\"Execute grounded PyAutoGUI code.\",\"name\":\"computer_use\",\"pa(...TRUNCATED) | [{"bytes":"iVBORw0KGgoAAAANSUhEUgAABQAAAALQCAIAAABAH0oBAAEAAElEQVR4nOx9d5xkVZX/OffeFyp07sk5MYmJDANDG(...TRUNCATED) | [] | [] | desktop_gui | en | Task Management & Collaboration | xlangai/AgentNet | d76ee50a63fad81cfdbe576416757d7c2091ed50 | raw | train | 20240930231654_2a922a80-5c58-4052-a49c-912cac533815 | https://huggingface.co/datasets/xlangai/AgentNet/tree/d76ee50a63fad81cfdbe576416757d7c2091ed50 | mit | 20240930231654_2a922a80-5c58-4052-a49c-912cac533815 | 8f60c320bae01534643436e8 | 1.428571 | true | human_annotation_metadata_ok_high_alignment_unambiguous_completed_all_steps_correct_nonredundant | false | [] | [
"canonicalized_without_sampling",
"quality_gated",
"deduplicated"
] | 4e3b4a65266e9671a272fda4bffe95487a9e5c91f44a7b67efe03b97ff976ed2 | 9233ba998a8be41f7f8cff7ab8ec5f144b0ce62e45e60f8e13b9e2e04dcdd8b8 | 9bc16f23aa157080029dd98ac5d61f56887967eefd1f98b6ec88cc8b242aa6b2 | "{\"source_config\":\"raw\",\"source_repo\":\"xlangai/AgentNet\",\"source_revision\":\"d76ee50a63fad(...TRUNCATED) | "{\"alignment_score\":10,\"applications\":[\"google chrome\"],\"difficulty\":6,\"efficiency_score\":(...TRUNCATED) | |
train-4cbbf3b76b6eb164aa9997a1 | train | computer_use_trajectory | [{"role":"system","content":[{"type":"text","text":"You are a cross-platform computer-use agent. Use(...TRUNCATED) | ["{\"function\":{\"description\":\"Execute grounded PyAutoGUI code.\",\"name\":\"computer_use\",\"pa(...TRUNCATED) | [{"bytes":"iVBORw0KGgoAAAANSUhEUgAABKgAAAMCCAIAAADCn2l0AAEAAElEQVR4nOy9d7xdVZU4vtbe+5TbX0+vkIRAigkld(...TRUNCATED) | [] | [] | desktop_gui | en | Education & Research | xlangai/AgentNet | d76ee50a63fad81cfdbe576416757d7c2091ed50 | raw | train | 20240924010603_766ac2ee-7503-4fb0-85a7-52c63482da3c | https://huggingface.co/datasets/xlangai/AgentNet/tree/d76ee50a63fad81cfdbe576416757d7c2091ed50 | mit | 20240924010603_766ac2ee-7503-4fb0-85a7-52c63482da3c | 4cbbf3b76b6eb164aa9997a1 | 1.428571 | true | human_annotation_metadata_ok_high_alignment_unambiguous_completed_all_steps_correct_nonredundant | false | [] | [
"canonicalized_without_sampling",
"quality_gated",
"deduplicated"
] | 02ed3d4cdedf95819a6ce765bc425d0d9cdab7b48500896c2c731249fecd839c | 28f6ff813b2a91250cd5e0ec7c01b56ae5d5aea2ae4025c6a4b5093967ae229d | 0f2277d58272162ca585aa0560ab242c8d26876743b71fe5f35220565c87fc8e | "{\"source_config\":\"raw\",\"source_repo\":\"xlangai/AgentNet\",\"source_revision\":\"d76ee50a63fad(...TRUNCATED) | "{\"alignment_score\":10,\"applications\":[\"google chrome\"],\"difficulty\":3,\"efficiency_score\":(...TRUNCATED) | |
train-619aa4d2fbf2c259d0d6173c | train | computer_use_trajectory | [{"role":"system","content":[{"type":"text","text":"You are a cross-platform computer-use agent. Use(...TRUNCATED) | ["{\"function\":{\"description\":\"Execute grounded PyAutoGUI code.\",\"name\":\"computer_use\",\"pa(...TRUNCATED) | [{"bytes":"iVBORw0KGgoAAAANSUhEUgAABQAAAALQCAIAAABAH0oBAAEAAElEQVR4nOydd5yUxf3HP1Oetru3dwdHryqCChJRs(...TRUNCATED) | [] | [] | desktop_gui | en | Office Tools | xlangai/AgentNet | d76ee50a63fad81cfdbe576416757d7c2091ed50 | raw | train | 20241002001023_0e7862f0-f89e-4b66-81e9-01f244cc4182 | https://huggingface.co/datasets/xlangai/AgentNet/tree/d76ee50a63fad81cfdbe576416757d7c2091ed50 | mit | 20241002001023_0e7862f0-f89e-4b66-81e9-01f244cc4182 | 619aa4d2fbf2c259d0d6173c | 1.285714 | true | human_annotation_metadata_ok_high_alignment_unambiguous_completed_all_steps_correct_nonredundant | false | [] | [
"canonicalized_without_sampling",
"quality_gated",
"deduplicated"
] | 15702a502de58da795a0c5c0552d35b3c71ab3c4dc3f4f4a260d4ad715c06033 | 088232c9541dd93245bd5e4d499937d75beea82daf9f3142094aafba5e8a97fe | f3c5c49a0c1857c6d49c22dbcfb5e2785896777f8519406cbd3f89063e7b2cb5 | "{\"source_config\":\"raw\",\"source_repo\":\"xlangai/AgentNet\",\"source_revision\":\"d76ee50a63fad(...TRUNCATED) | "{\"alignment_score\":10,\"applications\":[\"microsoft onenote\"],\"difficulty\":4,\"efficiency_scor(...TRUNCATED) | |
train-4348929a74ed4656e9fdc84e | train | computer_use_trajectory | [{"role":"system","content":[{"type":"text","text":"You are a cross-platform computer-use agent. Use(...TRUNCATED) | ["{\"function\":{\"description\":\"Execute grounded PyAutoGUI code.\",\"name\":\"computer_use\",\"pa(...TRUNCATED) | [{"bytes":"iVBORw0KGgoAAAANSUhEUgAABQAAAALQCAIAAABAH0oBAAEAAElEQVR4nOz9Z4Bcx3UgCp9TdVOn6YmYAGAwmEEGC(...TRUNCATED) | [] | [] | desktop_gui | en | News, Entertainment & Lifestyle | xlangai/AgentNet | d76ee50a63fad81cfdbe576416757d7c2091ed50 | raw | train | 20240924064340_c57793b1-cda7-4dc6-ba2b-3fbca1d7426c | https://huggingface.co/datasets/xlangai/AgentNet/tree/d76ee50a63fad81cfdbe576416757d7c2091ed50 | mit | 20240924064340_c57793b1-cda7-4dc6-ba2b-3fbca1d7426c | 4348929a74ed4656e9fdc84e | 1.428571 | true | human_annotation_metadata_ok_high_alignment_unambiguous_completed_all_steps_correct_nonredundant | false | [] | [
"canonicalized_without_sampling",
"quality_gated",
"deduplicated"
] | a1dd50ae68ecff558e29ba5c1edd2ca6c68f0c634a69e71253ed9c3a81661137 | 237a40f6cfec947bdcf6ca1f2933b952e3b2e1074068b514be7b1dc20aa40585 | 15f1198d69d747a7d20ee0d08444d5090cd91ff507b043caf1c70252c13fac8e | "{\"source_config\":\"raw\",\"source_repo\":\"xlangai/AgentNet\",\"source_revision\":\"d76ee50a63fad(...TRUNCATED) | "{\"alignment_score\":10,\"applications\":[],\"difficulty\":3,\"efficiency_score\":10,\"system\":\"D(...TRUNCATED) | |
train-a35de2176650cd13a365765e | train | computer_use_trajectory | [{"role":"system","content":[{"type":"text","text":"You are a cross-platform computer-use agent. Use(...TRUNCATED) | ["{\"function\":{\"description\":\"Execute grounded PyAutoGUI code.\",\"name\":\"computer_use\",\"pa(...TRUNCATED) | [{"bytes":"iVBORw0KGgoAAAANSUhEUgAABQAAAALQCAIAAABAH0oBAAEAAElEQVR4nOx9d5wcR5X/e1XVYcImrXLOWVawLFtyD(...TRUNCATED) | [] | [] | desktop_gui | en | Social Media & Communication | xlangai/AgentNet | d76ee50a63fad81cfdbe576416757d7c2091ed50 | raw | train | 20240927004810_46796b7f-d770-4a30-b7fc-2e3baa5aea89 | https://huggingface.co/datasets/xlangai/AgentNet/tree/d76ee50a63fad81cfdbe576416757d7c2091ed50 | mit | 20240927004810_46796b7f-d770-4a30-b7fc-2e3baa5aea89 | a35de2176650cd13a365765e | 1.428571 | true | human_annotation_metadata_ok_high_alignment_unambiguous_completed_all_steps_correct_nonredundant | false | [] | [
"canonicalized_without_sampling",
"quality_gated",
"deduplicated"
] | c9e65c6b34fd978ddd0041bfa513327dd58ce1acebdc5b48d7526eb835a3cd9b | 86eb64562a6a60dadb88f8962169ed92338a850830cb09648c3a7593d669b8cd | a5a4c8d5cf9a9cff91176a5014a9a90de0bdb13208355bb04ec37534b9a776e2 | "{\"source_config\":\"raw\",\"source_repo\":\"xlangai/AgentNet\",\"source_revision\":\"d76ee50a63fad(...TRUNCATED) | "{\"alignment_score\":10,\"applications\":[\"discord\"],\"difficulty\":5,\"efficiency_score\":10,\"s(...TRUNCATED) | |
train-3b3e7708b26a2f97922e6148 | train | computer_use_trajectory | [{"role":"system","content":[{"type":"text","text":"You are a cross-platform computer-use agent. Use(...TRUNCATED) | ["{\"function\":{\"description\":\"Execute grounded PyAutoGUI code.\",\"name\":\"computer_use\",\"pa(...TRUNCATED) | [{"bytes":"iVBORw0KGgoAAAANSUhEUgAABLwAAAL2CAIAAADErhhdAAEAAElEQVR4nOz9d5RVx5Uvju+qE26+nXMD3TSNmtDkI(...TRUNCATED) | [] | [] | desktop_gui | en | Development & Engineering | xlangai/AgentNet | d76ee50a63fad81cfdbe576416757d7c2091ed50 | raw | train | 20240927112736_b4fe0d7e-7b3b-4f28-aa24-3bde745d7453 | https://huggingface.co/datasets/xlangai/AgentNet/tree/d76ee50a63fad81cfdbe576416757d7c2091ed50 | mit | 20240927112736_b4fe0d7e-7b3b-4f28-aa24-3bde745d7453 | 3b3e7708b26a2f97922e6148 | 1.428571 | true | human_annotation_metadata_ok_high_alignment_unambiguous_completed_all_steps_correct_nonredundant | false | [] | [
"canonicalized_without_sampling",
"quality_gated",
"deduplicated"
] | 9f00ce29f0a68c28e2d0aecd1580da1d92e8ee97070e5b428820701bd69b46d3 | a7e8f188410dea374c612b2acbcf83fba6841267013d5e2146e53d5e0885c546 | 4ceee80730970eb56174c54bdd58c692f54ce637bf03a442b5e511be2ac1d723 | "{\"source_config\":\"raw\",\"source_repo\":\"xlangai/AgentNet\",\"source_revision\":\"d76ee50a63fad(...TRUNCATED) | "{\"alignment_score\":10,\"applications\":[\"remix ide\"],\"difficulty\":3,\"efficiency_score\":10,\(...TRUNCATED) |
End of preview. Expand in Data Studio
Qwen3.8 Executor Unified Training Dataset
Purpose
Public, provenance-pinned executor records for tool use, repository agents, computer use, and Korean coverage. This is an attributed integration; the upstream authors collected or generated the source data.
Schema
The Parquet files share the canonical fields documented in manifests/schema.json. messages uses role/content/tool-call structs. Dynamic tool arguments are validated JSON in arguments_json. Images and executable artifacts are self-contained bytes where included.
Sources, revisions, licenses, and transformations
| HF repo | revision | config | source split | license | rows before | accepted | rejected | transformation |
|---|---|---|---|---|---|---|---|---|
| tuandunghcmut/toolbench-v1 | 36de9b189753ad5de276181974f97df15e8c3202 |
default |
train |
apache-2.0 | 187542 | 181695 | 5847 | canonicalized |
| glaiveai/glaive-function-calling-v2 | e7f4b6456019f5d8bcb991ef0dd67d8ff23221ac |
default |
train |
apache-2.0 | 112960 | 46430 | 66530 | canonicalized |
| NousResearch/hermes-function-calling-v1 | dae3e1d28cfbcf4b915c04ea1e072030529b4bda |
func_calling_singleturn |
train |
apache-2.0 | 1893 | 1096 | 797 | canonicalized |
| NousResearch/hermes-function-calling-v1 | dae3e1d28cfbcf4b915c04ea1e072030529b4bda |
func_calling |
train |
apache-2.0 | 1893 | 1094 | 799 | canonicalized |
| NousResearch/hermes-function-calling-v1 | dae3e1d28cfbcf4b915c04ea1e072030529b4bda |
glaive_func_calling |
train |
apache-2.0 | 5209 | 3998 | 1211 | canonicalized |
| NousResearch/hermes-function-calling-v1 | dae3e1d28cfbcf4b915c04ea1e072030529b4bda |
json_mode_agentic |
train |
apache-2.0 | 1342 | 1284 | 58 | canonicalized |
| NousResearch/hermes-function-calling-v1 | dae3e1d28cfbcf4b915c04ea1e072030529b4bda |
json_mode_singleturn |
train |
apache-2.0 | 1241 | 1240 | 1 | canonicalized |
| heegyu/glaive-function-calling-v2-ko | c8b54e70bfbdd2fd6be037557e6feaf1abf14f5d |
default |
train |
apache-2.0 | 15170 | 15035 | 135 | canonicalized |
| nvidia/Open-SWE-Traces | ad4805a5aa7de70d99cab0bb8f99b15304c76de0 |
openhands |
minimax_m25 |
cc-by-4.0 | 49948 | 45165 | 4783 | canonicalized |
| nvidia/Open-SWE-Traces | ad4805a5aa7de70d99cab0bb8f99b15304c76de0 |
openhands |
qwen35_122b |
cc-by-4.0 | 55488 | 50364 | 5124 | canonicalized |
| nvidia/Open-SWE-Traces | ad4805a5aa7de70d99cab0bb8f99b15304c76de0 |
sweagent |
minimax_m25 |
cc-by-4.0 | 57268 | 53035 | 4233 | canonicalized |
| nvidia/Open-SWE-Traces | ad4805a5aa7de70d99cab0bb8f99b15304c76de0 |
sweagent |
qwen35_122b |
cc-by-4.0 | 44785 | 40375 | 4410 | canonicalized |
| nvidia/ProCUA-SFT | 120de7e954f851c2d24399230367f2b01ff815f9 |
raw |
shard_00000 |
cc-by-4.0 | 93566 | 1861 | 91705 | 1_of_50_deterministic_shards |
| nvidia/Nemotron-Personas-Korea | ada0f5b53a38bb5a30cce09358adde883c1ab63a |
default |
train |
cc-by-4.0 | 1000000 | 274531 | 725469 | deterministic_27.5_percent_sample |
| NousResearch/Hermes-3-Dataset | b1fddbdcae4e6714889365d1e6ce266a45289cc9 |
default |
train |
apache-2.0 | 958829 | 296777 | 662052 | executor_user_intent_filter_all_matches |
| mlx-community/ToolMind | 7b8e1d44d8dbe0e707dc336fdc1f7091fd6a9273 |
graphsyn |
train |
apache-2.0 | 163180 | 160849 | 2331 | complete_pinned_component_after_normalization_and_dedup |
| mlx-community/ToolMind | 7b8e1d44d8dbe0e707dc336fdc1f7091fd6a9273 |
APIGen-MT-5k-query |
train |
apache-2.0 | 25109 | 24571 | 538 | complete_pinned_component_after_normalization_and_dedup |
| mlx-community/ToolMind | 7b8e1d44d8dbe0e707dc336fdc1f7091fd6a9273 |
BUTTONInstruct-query |
train |
apache-2.0 | 21202 | 21136 | 66 | complete_pinned_component_after_normalization_and_dedup |
| mlx-community/ToolMind | 7b8e1d44d8dbe0e707dc336fdc1f7091fd6a9273 |
ToolACE-query |
train |
apache-2.0 | 7327 | 7307 | 20 | complete_pinned_component_after_normalization_and_dedup |
| mlx-community/ToolMind | 7b8e1d44d8dbe0e707dc336fdc1f7091fd6a9273 |
When2Call-query |
train |
apache-2.0 | 17531 | 17402 | 129 | complete_pinned_component_after_normalization_and_dedup |
| mlx-community/ToolMind | 7b8e1d44d8dbe0e707dc336fdc1f7091fd6a9273 |
glaive-function-calling-v2-query |
train |
apache-2.0 | 20017 | 19955 | 62 | complete_pinned_component_after_normalization_and_dedup |
| mlx-community/ToolMind | 7b8e1d44d8dbe0e707dc336fdc1f7091fd6a9273 |
tau-train-query |
train |
apache-2.0 | 12882 | 12879 | 3 | complete_pinned_component_after_normalization_and_dedup |
| mlx-community/ToolMind | 7b8e1d44d8dbe0e707dc336fdc1f7091fd6a9273 |
xlam-function-calling-60k-query |
train |
apache-2.0 | 101363 | 83785 | 17578 | complete_pinned_component_after_normalization_and_dedup |
| greghavens/fable-5-coding-and-debugging-traces | c63e82adec30798edcbd6e1dcb0014d2b15de236 |
default |
train |
cc-by-4.0 | 12490 | 2160 | 10330 | verified_final_train_trajectories_only; validation_and_prefix_views_excluded |
| Nanbeige/ToolMind-Web-QA | 2690dcdfdd82ab147aad4a65ab231d4e344f5cd0 |
open-wiki-traj |
train |
apache-2.0 | 5624 | 3859 | 1765 | answer_validated; >=2 evidence URLs; dated freshness anchor; benchmark markers excluded |
No single license is asserted over the mixture. Apply each row's source_license and upstream attribution requirements.
Build and isolation methodology
Evaluation data is canonicalized first. Exact SHA-256, normalized SHA-256 (whitespace, timestamps, UUIDs, and volatile IDs normalized), normalized user-prompt SHA-256, and 64-bit SimHash are then checked while admitting Train rows. Exact, normalized, prompt collisions, and near matches at Hamming distance <=3 are removed from Train. A trajectory/task is always one row and never split by turn.
High-confidence secrets, Korean resident registration numbers, missing user queries, empty assistant targets, invalid JSON tool arguments, missing tool references, broken images, and duplicate rows are rejected. persona_seed rows are excluded from trainable_rows. OpenWebRL and Uni-GUI are reference-only because their cards identify benchmark-derived task families. ProCUA uses deterministic shard 00000 only. Hermes-3 keeps executor-relevant rows; Korean personas use a deterministic 27.5% sample. ToolMind uses every pinned component after normalization and deduplication. Fable-5 keeps one complete verified Train trajectory and excludes cumulative prefix views and validation tasks. ToolMind-Web-QA requires upstream answer validation, at least two evidence URLs, a dated freshness anchor, and no known benchmark marker.
Multimodal and tool handling
Screenshots are not replaced with text descriptions. Included screenshots use bytes/path/mime/hash records and were decoded with Pillow. Tool definitions remain stable JSON strings; calls have name/id/type and validated arguments_json.
Recommended filters and limitations
Sample by record_type and source instead of uniform row sampling. Exclude persona_seed from direct response loss. Eval is not an independent benchmark where metadata.contamination_family says ToolBench. Source licenses and upstream terms still apply. The near-duplicate index uses a bounded within-Train collision bucket; exact and normalized checks are exhaustive.
Reproduction
pip install huggingface_hub pyarrow pillow requests numpy
python build_executor_datasets.py --output executor_dataset_build --cache /mnt/HDD0/executor_dataset_cache
python build_executor_datasets.py --output executor_dataset_build --cache /mnt/HDD0/executor_dataset_cache --eval-seed-repo snowman0919/qwen38-executor-eval-v1 --eval-seed-revision d348a8f0d951337f427f37cd027fffa3de26631f
python build_executor_datasets.py --validate-only --output executor_dataset_build
See manifests/, reports/, and reports/sanity_sample.jsonl for pinned source metadata and validation evidence.
V2 quality-only additions
V2 preserves every V1 Parquet file byte-for-byte through a Hub server-side repository copy and adds 90,251 canonical records. No row-count cap or size-based sampling was applied.
open-thoughts/OpenThoughts-114k@bd093c3994fd54d2390985b66988ddf282a55eb6: complete rows with non-empty problem, reasoning, generated solution, and ground-truth solution.markov-ai/computer-use@de58c88b4b33dd03fa4d5d0f490748f576bd37b3: score 1.0, every execution successful, no execution error, terminaldone=true, and all screenshots decodable.- OSWorld-derived contamination in
markov-ai/computer-useis intentional for practical executor performance; OSWorld evaluation is therefore not independent. markov-ai/cad-1000-hoursis not redistributed because the pinned repository declares no license.
See reports/v2_merge_report.json and merge_executor_dataset_v2.py for exact gates and revisions.
V2 reviewed candidate additions
Added 406,294 quality-gated, revision-pinned records from Ko-Agent, NextSearch (Apache config), SWE-smith resolved trajectories, Orchard GUI, UI-MOPD Desktop-2, and AgentNet. OSWorld/WebVoyager contamination is intentional for practical executor tuning, so those benchmarks are not independent. Exact gates and counts are in reports/v2_candidate_merge_report.json.
- Downloads last month
- 131
Models trained or fine-tuned on snowman0919/qwen38-executor-train-v2