bartowski/OS-Copilot_OS-Shepherd-9B-GGUF
Image-Text-to-Text • 9B • Updated • 2.59k
id stringlengths 48 138 | messages listlengths 3 3 | images listlengths 1 10 | coordinates listlengths 1 10 | results listlengths 1 3 |
|---|---|---|---|---|
000b951d-552d-5e6f-86a7-7500a32f86c2__first1last2__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/4d/1d/4d1d11640c2163fc1fc2084483771d4483d52355.png",
"osreward_rm_train_bundle/images/e4/15/e41565e0d94f415555c0d9d7a674642107a2a17a.png",
"osreward_rm_train_bundle/images/c1/b8/c1b828b3467eb51fff260c0ee127b3343acc011b.png"
] | [
null,
null,
null
] | [
{
"judge": "FAIL",
"setting": "first1last2",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 0,
"alignment_score": 0.5,
"efficiency": 0,
"self_correction": null
}
}
] |
002c609e-a8cf-5276-9e5d-73603edb14d7__first1last2__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/5f/5e/5f5e2426c8c2b081a0c489a81800f357aa112081.png",
"osreward_rm_train_bundle/images/28/5b/285be38810a80bb19162d09abc632ea50d469669.png",
"osreward_rm_train_bundle/images/8a/f5/8af58d7fe54b4f3d6cbf076272709a3ae954d36a.png"
] | [
[
46,
65
],
[
35,
62
],
null
] | [
{
"judge": "FAIL",
"setting": "first1last2",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 0,
"alignment_score": 0,
"efficiency": 0,
"self_correction": null
}
}
] |
00909a70-c086-5558-b3cd-9ec8ea5a1acd__last3__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/54/7a/547ac87a76d74a84f2118b29e2d97239a32f2a8b.png",
"osreward_rm_train_bundle/images/9f/db/9fdbb0f1c915de355178f28741ed2e4e5b451c35.png",
"osreward_rm_train_bundle/images/ca/fa/cafa2c33a5ca3b546aa4c37f355cbea830123525.png"
] | [
null,
[
158,
217
],
null
] | [
{
"judge": "FAIL",
"setting": "last3",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 0,
"alignment_score": 0.5,
"efficiency": 0,
"self_correction": null
}
}
] |
00e2e8ea-6187-5a17-bb20-8f46e8a1f861__last10__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/c4/9a/c49a3a2c3be6b332768f24de709dc48157ce9457.png",
"osreward_rm_train_bundle/images/05/55/0555cccfce7e923edf2a8cf7fa82646da0a5078d.png",
"osreward_rm_train_bundle/images/9d/d0/9dd04ca877017ee3cf781d588633873d25e1b56d.png",
"osreward_rm_train_bundle/images/64/65/6465ee37494e8... | [
[
72,
181
],
[
45,
95
],
[
205,
95
],
[
44,
95
],
[
394,
94
],
[
217,
94
],
[
859,
183
],
[
109,
148
],
[
91,
465
],
null
] | [
{
"judge": "FAIL",
"setting": "last10",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 0,
"alignment_score": 0,
"efficiency": 0,
"self_correction": null
}
}
] |
0168e1fc-5b66-502b-a504-284252fccf2a__last10__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/21/5e/215e83f32dbdb830346b55e7a328f2fc349b0f2f.png",
"osreward_rm_train_bundle/images/8e/45/8e454ca10d4d823b640c15ec5c08d07861a89be3.png",
"osreward_rm_train_bundle/images/fb/8a/fb8a8f19641d47a34fca6228b0130f5ced0d0624.png",
"osreward_rm_train_bundle/images/7f/2a/7f2adb319e246... | [
null,
null,
null,
null
] | [
{
"judge": "FAIL",
"setting": "last10",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 0,
"alignment_score": 0.5,
"efficiency": 0,
"self_correction": null
}
}
] |
017f33ff-9d5f-59a8-b222-c383a8bf4bbb__last10__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/62/7a/627aba98bbd3ad4cf88307e7e53b6e8fb038cf8e.png",
"osreward_rm_train_bundle/images/65/72/657236a34470c69340ce9a02aeab58cc9f9664e0.png",
"osreward_rm_train_bundle/images/59/41/5941b5e59ea3c7411c506a2e8f3f5474953eba6b.png",
"osreward_rm_train_bundle/images/79/a4/79a4caff75d17... | [
[
156,
625
],
null,
null,
[
147,
243
],
null
] | [
{
"judge": "SUCCESS",
"setting": "last10",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 1,
"alignment_score": 1,
"efficiency": 1,
"self_correction": null
}
}
] |
01806d13-2fb5-566d-81ee-578e364af3b8__first1last2__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/c7/f6/c7f687edb7d267b7353ecb3bb2f840892285978a.png",
"osreward_rm_train_bundle/images/39/bf/39bfd8d5ccf0fdd32d49efedfb493ec017d88a25.png",
"osreward_rm_train_bundle/images/93/0b/930baa436bf89950d755252b1e1b9dccb72ca984.png"
] | [
null,
null,
null
] | [
{
"judge": "SUCCESS",
"setting": "first1last2",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 1,
"alignment_score": 0.5,
"efficiency": 0.5,
"self_correction": null
}
}
] |
01993f57-730b-592e-a81c-2b4317a23c8b__last5__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/7e/d0/7ed06996ab3d027d81a8b2690ff19385886a054c.png",
"osreward_rm_train_bundle/images/ec/cb/eccb2d9098bda7ba4dfd37c89be2c893086ff64d.png",
"osreward_rm_train_bundle/images/4e/08/4e0805891e8ebf833b519f1bb3d482a4320c66f2.png",
"osreward_rm_train_bundle/images/7b/41/7b41136cce6ad... | [
null,
null,
null,
null
] | [
{
"judge": "FAIL",
"setting": "last5",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 0,
"alignment_score": 0.5,
"efficiency": 0,
"self_correction": null
}
}
] |
01a80eb3-3df5-59f2-a9da-bb7a8051a909__first1last2__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/96/dd/96ddcbf13a7b2c0440b928051697292f64ea8a7f.png",
"osreward_rm_train_bundle/images/28/d1/28d11aa488a36f0b36064273b3672ad44848bd5d.png",
"osreward_rm_train_bundle/images/cb/da/cbda7f45f1823a3f8488dc4d21014a89b6a645ff.png"
] | [
null,
[
194,
243
],
null
] | [
{
"judge": "SUCCESS",
"setting": "first1last2",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 1,
"alignment_score": 0.5,
"efficiency": 0.5,
"self_correction": null
}
}
] |
01eaf843-8189-5e62-ac49-9aba9e10333b__last5__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/12/31/1231da3ea7e5370c3580afe74ac76dc0d4f69747.png",
"osreward_rm_train_bundle/images/1a/e3/1ae3283c3f4956e2796cc2fb4c0e937a7bf189fa.png",
"osreward_rm_train_bundle/images/cb/49/cb4910bc55d83eed872e4f29f51a9164a9d5895d.png",
"osreward_rm_train_bundle/images/1c/91/1c91ecb03bf67... | [
[
194,
243
],
null,
[
103,
242
],
null
] | [
{
"judge": "SUCCESS",
"setting": "last5",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 1,
"alignment_score": 0.5,
"efficiency": 0.5,
"self_correction": null
}
}
] |
021acf2c-f8b7-5762-ae69-ae66c5bdc898__first1last2__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/1e/3e/1e3eaa5fd35a63540cd3698f6445a1fccb305a87.png",
"osreward_rm_train_bundle/images/cd/7e/cd7e9ec38593f00025b0cbc86aa66f5f8bc8819e.png",
"osreward_rm_train_bundle/images/b5/3e/b53ec69e23cce11a54ff94b570e326093f9003fb.png"
] | [
[
42,
65
],
[
234,
344
],
null
] | [
{
"judge": "SUCCESS",
"setting": "first1last2",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 1,
"alignment_score": 1,
"efficiency": 0.5,
"self_correction": null
}
}
] |
026e3363-c65d-5fef-9ceb-3043a2de119e__first1last2__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/b4/b9/b4b9f6c365f3ef3c4434cfbfa3995ee85eaab96c.png",
"osreward_rm_train_bundle/images/82/82/82823dcfef2602bc320550a3d7edf7bde9a9a065.png",
"osreward_rm_train_bundle/images/d0/f2/d0f20a4c322dc8c4a3e322578ccbea9b971f541f.png"
] | [
[
175,
69
],
[
441,
470
],
null
] | [
{
"judge": "FAIL",
"setting": "first1last2",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 0,
"alignment_score": 0.5,
"efficiency": 0,
"self_correction": null
}
}
] |
030c40a7-4d05-54b4-b358-3f0eff224c6f__first1last2__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/5b/1c/5b1cea7208c6839a52bfd677a1d7538274f72eec.png",
"osreward_rm_train_bundle/images/ca/39/ca390a47f9a5623ea66914e86d1990b959b96631.png",
"osreward_rm_train_bundle/images/61/cb/61cbc4d8e1ad4d6690b956ba1539fd6f1dbfd5c8.png"
] | [
[
115,
65
],
[
85,
270
],
null
] | [
{
"judge": "FAIL",
"setting": "first1last2",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 0,
"alignment_score": 0.5,
"efficiency": 0,
"self_correction": null
}
}
] |
031845e0-bbf2-57e7-b66f-ec8e620c38fe__last10__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/59/c9/59c969cd1a086756fa11a390100f595e8c3a08bb.png",
"osreward_rm_train_bundle/images/ec/b5/ecb552d8039a7428f98649788df4f8d09f882349.png",
"osreward_rm_train_bundle/images/9c/57/9c579ca2a7377cf3935d9d15962eb6dbad573663.png",
"osreward_rm_train_bundle/images/ee/3f/ee3fcefb07b45... | [
[
77,
11
],
[
94,
195
],
[
547,
415
],
[
490,
414
],
[
500,
414
],
null,
null,
null
] | [
{
"judge": "SUCCESS",
"setting": "last10",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 1,
"alignment_score": 0.5,
"efficiency": 0.5,
"self_correction": null
}
}
] |
0351105d-68c8-5280-a714-d477e80516e6__first1last2__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/36/c4/36c46d2a89908cd3b02ca98b767e6d5cdf284e45.png",
"osreward_rm_train_bundle/images/45/fe/45fedb7d8db1381a5a065cebd00ae50d91230ef9.png",
"osreward_rm_train_bundle/images/48/1c/481ce7640cefe7f28c748a8ea0b679680595099b.png"
] | [
[
104,
242
],
[
394,
636
],
null
] | [
{
"judge": "FAIL",
"setting": "first1last2",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 0,
"alignment_score": 0.5,
"efficiency": 0,
"self_correction": null
}
}
] |
037bab4e-9cbf-5cc2-977e-d298f2a01f14__first1last2__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/e3/d1/e3d1c0cf275898b642aa4cab9fd1f69528e03da0.png",
"osreward_rm_train_bundle/images/7e/fa/7efaa5639d60a2a25c2c0e3b7a2aa2e4326fc90e.png",
"osreward_rm_train_bundle/images/0e/03/0e03e953a1c1cbfb0dc31d6be4faa53fb8dd8353.png"
] | [
[
258,
87
],
null,
null
] | [
{
"judge": "SUCCESS",
"setting": "first1last2",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 1,
"alignment_score": 0.5,
"efficiency": 0.5,
"self_correction": null
}
}
] |
039f6bfa-e4da-585b-8dab-6309197cb1c0__last10__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/5d/1e/5d1e7787e4d34f9669ce0c7beeb4d437f310a09b.png",
"osreward_rm_train_bundle/images/86/4d/864d1f54731f959949697ce8859ec1c9c1c13841.png",
"osreward_rm_train_bundle/images/27/b8/27b89119df76de3dd11c03dd2fa5706c6b850be4.png",
"osreward_rm_train_bundle/images/64/d3/64d3b89f72ac2... | [
[
104,
242
],
null,
[
393,
636
],
null,
null
] | [
{
"judge": "SUCCESS",
"setting": "last10",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 1,
"alignment_score": 1,
"efficiency": 1,
"self_correction": null
}
}
] |
03a24761-20d7-51b9-90fc-0c1a532aae67__last5__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/6e/3d/6e3de7a735a834287c7717ab154ae3871a75e207.png",
"osreward_rm_train_bundle/images/64/bb/64bb6ad930380021dbaa1c679c78a460147e73a9.png",
"osreward_rm_train_bundle/images/9f/ab/9fab7ba1c5c4b709e9dac8b877615724a327f3d5.png",
"osreward_rm_train_bundle/images/ad/86/ad8636b01f10d... | [
[
46,
65
],
[
78,
231
],
[
632,
698
],
null
] | [
{
"judge": "SUCCESS",
"setting": "last5",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 1,
"alignment_score": 1,
"efficiency": 1,
"self_correction": null
}
}
] |
03af81d7-8866-5107-95d9-621fddededbb__last10__gemini-3.1-pro-preview | [
{
"role": "system",
"content": "[ROLE]\nYou are a reward model evaluating a GUI agent operating across diverse platforms (e.g., Desktop, Web, Mobile). Your job is to determine whether the agent's trajectory successfully completes the user's task.\n\n[INPUTS]\nYou will receive varying combinations of the fol... | [
"osreward_rm_train_bundle/images/01/e6/01e632eb3d1084187220fe0639d9b12405f505fd.png",
"osreward_rm_train_bundle/images/96/35/9635cde85a9971861b53252540de4e9186ac69f3.png",
"osreward_rm_train_bundle/images/b3/00/b3007419cf99f8c82e5152d97644b00703dc624f.png",
"osreward_rm_train_bundle/images/ae/04/ae048ff7b6760... | [
[
52,
148
],
[
825,
262
],
[
86,
323
],
null
] | [
{
"judge": "SUCCESS",
"setting": "last10",
"model": "gemini-3.1-pro-preview",
"metrics": {
"success": 1,
"alignment_score": 0.5,
"efficiency": 0.5,
"self_correction": null
}
}
] |
End of preview. Expand in Data Studio
OSReward Data
OSReward is a multimodal reward-model dataset for judging whether GUI-agent trajectories complete the user's task. The release contains:
- Supervised fine-tuning data in a LLaMAFactory-compatible ShareGPT layout.
- Deduplicated screenshots packaged in independently extractable tar shards.
- Training and validation data for GRPO.
The expected answer contains a brief evidence-based analysis followed by a final line in exactly one of these forms:
Judge: SUCCESS
Judge: FAIL
Layout
sft/datasets/ SFT JSON files
sft/viewer/ Parquet view of the SFT data for Hub preview
sft/images-shards/ independent image tar shards
rl/train.parquet GRPO training split
rl/val.parquet GRPO validation split
stats/ compact release statistics
Quick Start
Extract all screenshot shards from the repository root:
for shard in sft/images-shards/*.tar; do tar -xf "$shard"; done
This restores paths under osreward_rm_train_bundle/images/, matching the
relative image paths used by both SFT and RL records.
The SFT JSON files are the canonical LLaMAFactory training files. The Parquet
files under sft/viewer/ contain the same records in a format supported by
the Hub dataset preview.
See DATA_CARD.md, sft/SFT_FORMAT.md, and rl/RL_FORMAT.md for schemas and limitations.
Integrity
Verify the downloaded release with:
sha256sum -c SHA256SUMS
No model weights or training framework checkout is included.
WebTrail 100K completion increment
The release now contains exactly 100,000 SFT judge instances. The additive
WebTrail subset contributes 6,386 unique web trajectories using the last up to
five screenshots (last5); trajectories shorter than five states naturally
contain fewer images. Its canonical files are
sft/datasets/osreward_webtrail_supplement_single_filter.json and
sft/viewer/osreward_webtrail_supplement_single_filter.parquet.
The 32 independently extractable WebTrail image shards are described by
sft/webtrail_image_shards_manifest.json. The existing extraction loop over
sft/images-shards/*.tar includes both the original and WebTrail shards.
- Downloads last month
- 579
Models trained or fine-tuned on OS-Copilot/OS-Shepherd-100K
