{ "schema_version": 1, "title": "Repro - Enhancing Reasoning for Diffusion LLMs via Distribution Matching Policy Optimization", "emoji": "🎯", "space_id": "batesm/repro-dmpo", "paper": { "arxiv_id": "2510.08233" }, "tags": [ "icml2026-repro", "paper-09CSjVeDug" ], "updated_at": "2026-07-15T16:54:55+00:00", "root": { "slug": "index", "title": "Repro - Enhancing Reasoning for Diffusion LLMs via Distribution Matching Policy Optimization", "file": "pages/index.md", "children": [ { "slug": "claim-1-dmpo-distribution-matching-via-cross-entropy", "title": "Claim 1: DMPO distribution matching via cross-entropy", "file": "pages/claim-1-dmpo-distribution-matching-via-cross-entropy/page.md", "children": [] }, { "slug": "claim-2-weight-baseline-subtraction-for-small-batch-training", "title": "Claim 2: Weight baseline subtraction for small-batch training", "file": "pages/claim-2-weight-baseline-subtraction-for-small-batch-training/page.md", "children": [] }, { "slug": "claim-3-r1-zero-like-recipe-without-sft", "title": "Claim 3: R1-Zero-like recipe without SFT", "file": "pages/claim-3-r1-zero-like-recipe-without-sft/page.md", "children": [] }, { "slug": "claim-4-benchmark-comparison-against-baselines", "title": "Claim 4: Benchmark comparison against baselines", "file": "pages/claim-4-benchmark-comparison-against-baselines/page.md", "children": [] }, { "slug": "claim-5-accuracy-gains-up-to-54-3-over-sota", "title": "Claim 5: Accuracy gains up to 54.3% over SOTA", "file": "pages/claim-5-accuracy-gains-up-to-54-3-over-sota/page.md", "children": [] }, { "slug": "conclusion", "title": "Conclusion", "file": "pages/conclusion/page.md", "children": [] } ] }, "agent_view_tokens": 3747, "revision": "1784134495250998000" }