File size: 3,550 Bytes
589b85d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 | {
"schema_version": 2,
"title": "Reproduction: MASH: Modeling Abstention via Selective Help-Seeking",
"emoji": "🎯",
"space_id": "SabaPivot/repro-mash-selective-help-seeking",
"paper": {
"arxiv_id": "2510.01152"
},
"tags": [
"icml2026-repro",
"paper-d5rd1zlHTf"
],
"updated_at": "2026-07-28T14:49:50+00:00",
"root": {
"slug": "index",
"title": "Reproduction: MASH: Modeling Abstention via Selective Help-Seeking",
"file": "pages/index.md",
"children": [
{
"slug": "executive-summary",
"title": "Executive summary",
"file": "pages/executive-summary/page.md",
"children": []
},
{
"slug": "claim-1-on-hotpotqa-mash-improves-answer-accuracy-by-7-6-percentage-points",
"title": "Claim 1: On HotPotQA, MASH improves answer accuracy by 7.6 percentage points",
"file": "pages/claim-1-on-hotpotqa-mash-improves-answer-accuracy-by-7-6-percentage-points/page.md",
"children": []
},
{
"slug": "claim-2-on-2wikimultihopqa-mash-improves-accuracy-by-6-64-percentage-points-over",
"title": "Claim 2: On 2WikiMultihopQA, MASH improves accuracy by 6.64 percentage points over",
"file": "pages/claim-2-on-2wikimultihopqa-mash-improves-accuracy-by-6-64-percentage-points-over/page.md",
"children": []
},
{
"slug": "claim-3-mash-improves-average-tool-productivity-by-3-83-points-across-evaluated",
"title": "Claim 3: MASH improves average Tool Productivity by 3.83 points across evaluated",
"file": "pages/claim-3-mash-improves-average-tool-productivity-by-3-83-points-across-evaluated/page.md",
"children": []
},
{
"slug": "claim-4-on-the-subset-of-hotpotqa-questions-requiring-exactly-one-search",
"title": "Claim 4: On the subset of HotPotQA questions requiring exactly one search",
"file": "pages/claim-4-on-the-subset-of-hotpotqa-questions-requiring-exactly-one-search/page.md",
"children": []
},
{
"slug": "claim-5-in-the-abstention-inference-mode-mash-with-otc-strict-reaches-59-9",
"title": "Claim 5: In the abstention inference mode, MASH with OTC-Strict reaches 59.9%",
"file": "pages/claim-5-in-the-abstention-inference-mode-mash-with-otc-strict-reaches-59-9/page.md",
"children": []
},
{
"slug": "claim-6-mash-is-trained-via-a-two-step-pipeline-of-warm-start-supervised",
"title": "Claim 6: MASH is trained via a two-step pipeline of warm-start supervised",
"file": "pages/claim-6-mash-is-trained-via-a-two-step-pipeline-of-warm-start-supervised/page.md",
"children": []
},
{
"slug": "conclusion",
"title": "Conclusion",
"file": "pages/conclusion/page.md",
"children": []
}
]
},
"traces": [],
"workspace": {
"file": "workspace.json",
"file_count": 1,
"total_size": 23331,
"bucket_id": "SabaPivot/repro-mash-selective-help-seeking-artifacts"
},
"agent_view_tokens": 3800,
"trace_view_tokens": 10,
"workspace_view_tokens": 18,
"revision": "b6ce1f04feac8570c199",
"workspace_ref": {
"repo_id": "SabaPivot/repro-mash-selective-help-seeking-artifacts",
"repo_type": "bucket",
"repo_url": "https://huggingface.co/buckets/SabaPivot/repro-mash-selective-help-seeking-artifacts",
"private": false
},
"workspace_bucket": "https://huggingface.co/buckets/SabaPivot/repro-mash-selective-help-seeking-artifacts"
} |