{"adoption": {"forks": 26, "observed_at": "2026-08-28T04:03:35.399037+00:00", "stars": 1100}, "canonical_url": "https://ross.abutalabs.com/products/dft", "card": {"archived": false, "artifact_type": "library", "description": "[ICLR 2026] On the Generalization of SFT: A Reinforcement Learning Perspective with Reward Rectification.", "domain": ["large-language-models", "machine-learning", "deep-learning"], "enriched": true, "function": ["llm-training", "machine-learning", "deep-learning"], "health_score": 77, "homepage": "https://arxiv.org/abs/2508.05629", "language": "Python", "license": null, "license_family": "other", "maturity": "active", "member_repos": ["yongliang-wu/DFT"], "name": "yongliang-wu/DFT", "platform": ["python"], "pushed_at": "2026-08-01T04:15:51+00:00", "repo": "yongliang-wu/DFT", "stars": 1100, "tags": ["sft", "fine-tuning", "reinforcement-learning", "research-paper", "iclr-2026", "reward-rectification", "math-reasoning", "gpu"], "topics": [], "urls": [], "use_cases": ["fine-tune an LLM with better generalization than standard SFT", "improve supervised fine-tuning for math reasoning tasks", "apply a one-line loss modification to an existing SFT pipeline", "use DFT as a streamlined alternative to offline RL fine-tuning", "reproduce ICLR 2026 paper results on models up to 7B parameters"], "what_it_is": "DFT (Dynamic Fine-Tuning) is the official implementation of an ICLR 2026 paper that improves Supervised Fine-Tuning of LLMs by dynamically rescaling the loss with token probabilities, rectifying the implicit reward structure of SFT gradients. It requires only a single-line change to standard SFT and improves generalization on math reasoning, code generation, and multi-modal benchmarks.", "when_to_avoid": ["your fine-tuning domain is literary, financial, or otherwise outside math/code tasks where DFT is unproven", "you need a production-grade, licensed, well-supported training framework", "you work with models far beyond 7B parameters, which the paper did not evaluate"], "when_to_choose": ["you are fine-tuning LLMs on math or code datasets and want better generalization than vanilla SFT", "you want an RL-like generalization boost without running full RL training", "you are doing research on SFT vs RL training dynamics"]}, "data_as_of": "2026-08-30T08:39:29.467469+00:00", "members": [{"path": "/products/dft", "repo": "yongliang-wu/DFT", "role": "main", "score": 61}], "provenance": {"archived": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "artifact_type": {"confidence": null, "enriched_at": "2026-08-30T06:45:56.429077+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "c572b7b0f3690f149228c59a635da2aa422ba0f0b94efa83339da7cea0530ee3", "fetched_at": "2026-08-28T04:03:35.399037+00:00", "kind": "readme", "missing": false, "url": "https://github.com/yongliang-wu/DFT"}, {"content_hash": "6b1fa346495cfb92872e52d80d32e18bdba226f2d7f0444781a0e8bf7ed2c0ec", "fetched_at": "2026-08-29T12:49:27.802032+00:00", "kind": "homepage", "missing": false, "url": "https://arxiv.org/abs/2508.05629"}, {"content_hash": "47cbc55ff1deaca3f36b463373de863a7053d91cacf9f95725618e11610209ae", "fetched_at": "2026-08-29T12:49:27.810955+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/ourmembers.html"}, {"content_hash": "cca9c3a11c56b1212864e03fb8a5511a3bdb6df53398ba4dd2ed778f931607dd", "fetched_at": "2026-08-29T12:49:27.806705+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/donate.html"}, {"content_hash": "a1f16f915a9ad89ee57c3e929a1b6577d481f52c05f8015fc720b16ee8014f19", "fetched_at": "2026-08-29T12:49:27.813288+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about"}, {"content_hash": "b14a8d05a0ec337409ab63bffd91306b5317b16fac358aa6d1fe6afbaf3c8628", "fetched_at": "2026-08-29T12:49:27.809010+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/labs/index.html"}], "taxonomy_version": 1}, "description": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "domain": {"confidence": null, "enriched_at": "2026-08-30T06:45:56.429077+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "c572b7b0f3690f149228c59a635da2aa422ba0f0b94efa83339da7cea0530ee3", "fetched_at": "2026-08-28T04:03:35.399037+00:00", "kind": "readme", "missing": false, "url": "https://github.com/yongliang-wu/DFT"}, {"content_hash": "6b1fa346495cfb92872e52d80d32e18bdba226f2d7f0444781a0e8bf7ed2c0ec", "fetched_at": "2026-08-29T12:49:27.802032+00:00", "kind": "homepage", "missing": false, "url": "https://arxiv.org/abs/2508.05629"}, {"content_hash": "47cbc55ff1deaca3f36b463373de863a7053d91cacf9f95725618e11610209ae", "fetched_at": "2026-08-29T12:49:27.810955+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/ourmembers.html"}, {"content_hash": "cca9c3a11c56b1212864e03fb8a5511a3bdb6df53398ba4dd2ed778f931607dd", "fetched_at": "2026-08-29T12:49:27.806705+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/donate.html"}, {"content_hash": "a1f16f915a9ad89ee57c3e929a1b6577d481f52c05f8015fc720b16ee8014f19", "fetched_at": "2026-08-29T12:49:27.813288+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about"}, {"content_hash": "b14a8d05a0ec337409ab63bffd91306b5317b16fac358aa6d1fe6afbaf3c8628", "fetched_at": "2026-08-29T12:49:27.809010+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/labs/index.html"}], "taxonomy_version": 1}, "enriched": {"inputs": [], "kind": "computed", "method": "enrichment_status"}, "function": {"confidence": null, "enriched_at": "2026-08-30T06:45:56.429077+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "c572b7b0f3690f149228c59a635da2aa422ba0f0b94efa83339da7cea0530ee3", "fetched_at": "2026-08-28T04:03:35.399037+00:00", "kind": "readme", "missing": false, "url": "https://github.com/yongliang-wu/DFT"}, {"content_hash": "6b1fa346495cfb92872e52d80d32e18bdba226f2d7f0444781a0e8bf7ed2c0ec", "fetched_at": "2026-08-29T12:49:27.802032+00:00", "kind": "homepage", "missing": false, "url": "https://arxiv.org/abs/2508.05629"}, {"content_hash": "47cbc55ff1deaca3f36b463373de863a7053d91cacf9f95725618e11610209ae", "fetched_at": "2026-08-29T12:49:27.810955+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/ourmembers.html"}, {"content_hash": "cca9c3a11c56b1212864e03fb8a5511a3bdb6df53398ba4dd2ed778f931607dd", "fetched_at": "2026-08-29T12:49:27.806705+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/donate.html"}, {"content_hash": "a1f16f915a9ad89ee57c3e929a1b6577d481f52c05f8015fc720b16ee8014f19", "fetched_at": "2026-08-29T12:49:27.813288+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about"}, {"content_hash": "b14a8d05a0ec337409ab63bffd91306b5317b16fac358aa6d1fe6afbaf3c8628", "fetched_at": "2026-08-29T12:49:27.809010+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/labs/index.html"}], "taxonomy_version": 1}, "health_score": {"inputs": ["days_since_push", "days_since_release", "archived"], "kind": "computed", "method": "health_v1"}, "homepage": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "language": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "license": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "license_family": {"inputs": ["license"], "kind": "computed", "method": "license_family"}, "maturity": {"confidence": null, "enriched_at": "2026-08-30T06:45:56.429077+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "c572b7b0f3690f149228c59a635da2aa422ba0f0b94efa83339da7cea0530ee3", "fetched_at": "2026-08-28T04:03:35.399037+00:00", "kind": "readme", "missing": false, "url": "https://github.com/yongliang-wu/DFT"}, {"content_hash": "6b1fa346495cfb92872e52d80d32e18bdba226f2d7f0444781a0e8bf7ed2c0ec", "fetched_at": "2026-08-29T12:49:27.802032+00:00", "kind": "homepage", "missing": false, "url": "https://arxiv.org/abs/2508.05629"}, {"content_hash": "47cbc55ff1deaca3f36b463373de863a7053d91cacf9f95725618e11610209ae", "fetched_at": "2026-08-29T12:49:27.810955+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/ourmembers.html"}, {"content_hash": "cca9c3a11c56b1212864e03fb8a5511a3bdb6df53398ba4dd2ed778f931607dd", "fetched_at": "2026-08-29T12:49:27.806705+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/donate.html"}, {"content_hash": "a1f16f915a9ad89ee57c3e929a1b6577d481f52c05f8015fc720b16ee8014f19", "fetched_at": "2026-08-29T12:49:27.813288+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about"}, {"content_hash": "b14a8d05a0ec337409ab63bffd91306b5317b16fac358aa6d1fe6afbaf3c8628", "fetched_at": "2026-08-29T12:49:27.809010+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/labs/index.html"}], "taxonomy_version": 1}, "member_repos": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "name": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "platform": {"confidence": null, "enriched_at": "2026-08-30T06:45:56.429077+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "c572b7b0f3690f149228c59a635da2aa422ba0f0b94efa83339da7cea0530ee3", "fetched_at": "2026-08-28T04:03:35.399037+00:00", "kind": "readme", "missing": false, "url": "https://github.com/yongliang-wu/DFT"}, {"content_hash": "6b1fa346495cfb92872e52d80d32e18bdba226f2d7f0444781a0e8bf7ed2c0ec", "fetched_at": "2026-08-29T12:49:27.802032+00:00", "kind": "homepage", "missing": false, "url": "https://arxiv.org/abs/2508.05629"}, {"content_hash": "47cbc55ff1deaca3f36b463373de863a7053d91cacf9f95725618e11610209ae", "fetched_at": "2026-08-29T12:49:27.810955+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/ourmembers.html"}, {"content_hash": "cca9c3a11c56b1212864e03fb8a5511a3bdb6df53398ba4dd2ed778f931607dd", "fetched_at": "2026-08-29T12:49:27.806705+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/donate.html"}, {"content_hash": "a1f16f915a9ad89ee57c3e929a1b6577d481f52c05f8015fc720b16ee8014f19", "fetched_at": "2026-08-29T12:49:27.813288+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about"}, {"content_hash": "b14a8d05a0ec337409ab63bffd91306b5317b16fac358aa6d1fe6afbaf3c8628", "fetched_at": "2026-08-29T12:49:27.809010+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/labs/index.html"}], "taxonomy_version": 1}, "pushed_at": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "repo": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "stars": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "tags": {"confidence": null, "enriched_at": "2026-08-30T06:45:56.429077+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "c572b7b0f3690f149228c59a635da2aa422ba0f0b94efa83339da7cea0530ee3", "fetched_at": "2026-08-28T04:03:35.399037+00:00", "kind": "readme", "missing": false, "url": "https://github.com/yongliang-wu/DFT"}, {"content_hash": "6b1fa346495cfb92872e52d80d32e18bdba226f2d7f0444781a0e8bf7ed2c0ec", "fetched_at": "2026-08-29T12:49:27.802032+00:00", "kind": "homepage", "missing": false, "url": "https://arxiv.org/abs/2508.05629"}, {"content_hash": "47cbc55ff1deaca3f36b463373de863a7053d91cacf9f95725618e11610209ae", "fetched_at": "2026-08-29T12:49:27.810955+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/ourmembers.html"}, {"content_hash": "cca9c3a11c56b1212864e03fb8a5511a3bdb6df53398ba4dd2ed778f931607dd", "fetched_at": "2026-08-29T12:49:27.806705+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/donate.html"}, {"content_hash": "a1f16f915a9ad89ee57c3e929a1b6577d481f52c05f8015fc720b16ee8014f19", "fetched_at": "2026-08-29T12:49:27.813288+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about"}, {"content_hash": "b14a8d05a0ec337409ab63bffd91306b5317b16fac358aa6d1fe6afbaf3c8628", "fetched_at": "2026-08-29T12:49:27.809010+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/labs/index.html"}], "taxonomy_version": 1}, "topics": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "urls": {"kind": "observed", "observed_at": "2026-08-28T04:03:35.399037+00:00", "source": "github"}, "use_cases": {"confidence": null, "enriched_at": "2026-08-30T06:45:56.429077+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "c572b7b0f3690f149228c59a635da2aa422ba0f0b94efa83339da7cea0530ee3", "fetched_at": "2026-08-28T04:03:35.399037+00:00", "kind": "readme", "missing": false, "url": "https://github.com/yongliang-wu/DFT"}, {"content_hash": "6b1fa346495cfb92872e52d80d32e18bdba226f2d7f0444781a0e8bf7ed2c0ec", "fetched_at": "2026-08-29T12:49:27.802032+00:00", "kind": "homepage", "missing": false, "url": "https://arxiv.org/abs/2508.05629"}, {"content_hash": "47cbc55ff1deaca3f36b463373de863a7053d91cacf9f95725618e11610209ae", "fetched_at": "2026-08-29T12:49:27.810955+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/ourmembers.html"}, {"content_hash": "cca9c3a11c56b1212864e03fb8a5511a3bdb6df53398ba4dd2ed778f931607dd", "fetched_at": "2026-08-29T12:49:27.806705+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/donate.html"}, {"content_hash": "a1f16f915a9ad89ee57c3e929a1b6577d481f52c05f8015fc720b16ee8014f19", "fetched_at": "2026-08-29T12:49:27.813288+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about"}, {"content_hash": "b14a8d05a0ec337409ab63bffd91306b5317b16fac358aa6d1fe6afbaf3c8628", "fetched_at": "2026-08-29T12:49:27.809010+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/labs/index.html"}], "taxonomy_version": 1}, "what_it_is": {"confidence": null, "enriched_at": "2026-08-30T06:45:56.429077+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "c572b7b0f3690f149228c59a635da2aa422ba0f0b94efa83339da7cea0530ee3", "fetched_at": "2026-08-28T04:03:35.399037+00:00", "kind": "readme", "missing": false, "url": "https://github.com/yongliang-wu/DFT"}, {"content_hash": "6b1fa346495cfb92872e52d80d32e18bdba226f2d7f0444781a0e8bf7ed2c0ec", "fetched_at": "2026-08-29T12:49:27.802032+00:00", "kind": "homepage", "missing": false, "url": "https://arxiv.org/abs/2508.05629"}, {"content_hash": "47cbc55ff1deaca3f36b463373de863a7053d91cacf9f95725618e11610209ae", "fetched_at": "2026-08-29T12:49:27.810955+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/ourmembers.html"}, {"content_hash": "cca9c3a11c56b1212864e03fb8a5511a3bdb6df53398ba4dd2ed778f931607dd", "fetched_at": "2026-08-29T12:49:27.806705+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/donate.html"}, {"content_hash": "a1f16f915a9ad89ee57c3e929a1b6577d481f52c05f8015fc720b16ee8014f19", "fetched_at": "2026-08-29T12:49:27.813288+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about"}, {"content_hash": "b14a8d05a0ec337409ab63bffd91306b5317b16fac358aa6d1fe6afbaf3c8628", "fetched_at": "2026-08-29T12:49:27.809010+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/labs/index.html"}], "taxonomy_version": 1}, "when_to_avoid": {"confidence": null, "enriched_at": "2026-08-30T06:45:56.429077+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "c572b7b0f3690f149228c59a635da2aa422ba0f0b94efa83339da7cea0530ee3", "fetched_at": "2026-08-28T04:03:35.399037+00:00", "kind": "readme", "missing": false, "url": "https://github.com/yongliang-wu/DFT"}, {"content_hash": "6b1fa346495cfb92872e52d80d32e18bdba226f2d7f0444781a0e8bf7ed2c0ec", "fetched_at": "2026-08-29T12:49:27.802032+00:00", "kind": "homepage", "missing": false, "url": "https://arxiv.org/abs/2508.05629"}, {"content_hash": "47cbc55ff1deaca3f36b463373de863a7053d91cacf9f95725618e11610209ae", "fetched_at": "2026-08-29T12:49:27.810955+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/ourmembers.html"}, {"content_hash": "cca9c3a11c56b1212864e03fb8a5511a3bdb6df53398ba4dd2ed778f931607dd", "fetched_at": "2026-08-29T12:49:27.806705+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/donate.html"}, {"content_hash": "a1f16f915a9ad89ee57c3e929a1b6577d481f52c05f8015fc720b16ee8014f19", "fetched_at": "2026-08-29T12:49:27.813288+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about"}, {"content_hash": "b14a8d05a0ec337409ab63bffd91306b5317b16fac358aa6d1fe6afbaf3c8628", "fetched_at": "2026-08-29T12:49:27.809010+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/labs/index.html"}], "taxonomy_version": 1}, "when_to_choose": {"confidence": null, "enriched_at": "2026-08-30T06:45:56.429077+00:00", "kind": "inferred", "prompt_version": 1, "sources": [{"content_hash": "c572b7b0f3690f149228c59a635da2aa422ba0f0b94efa83339da7cea0530ee3", "fetched_at": "2026-08-28T04:03:35.399037+00:00", "kind": "readme", "missing": false, "url": "https://github.com/yongliang-wu/DFT"}, {"content_hash": "6b1fa346495cfb92872e52d80d32e18bdba226f2d7f0444781a0e8bf7ed2c0ec", "fetched_at": "2026-08-29T12:49:27.802032+00:00", "kind": "homepage", "missing": false, "url": "https://arxiv.org/abs/2508.05629"}, {"content_hash": "47cbc55ff1deaca3f36b463373de863a7053d91cacf9f95725618e11610209ae", "fetched_at": "2026-08-29T12:49:27.810955+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/ourmembers.html"}, {"content_hash": "cca9c3a11c56b1212864e03fb8a5511a3bdb6df53398ba4dd2ed778f931607dd", "fetched_at": "2026-08-29T12:49:27.806705+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about/donate.html"}, {"content_hash": "a1f16f915a9ad89ee57c3e929a1b6577d481f52c05f8015fc720b16ee8014f19", "fetched_at": "2026-08-29T12:49:27.813288+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/about"}, {"content_hash": "b14a8d05a0ec337409ab63bffd91306b5317b16fac358aa6d1fe6afbaf3c8628", "fetched_at": "2026-08-29T12:49:27.809010+00:00", "kind": "site_page", "missing": false, "url": "https://info.arxiv.org/labs/index.html"}], "taxonomy_version": 1}}, "score": {"components": {"activity": 95, "longevity": 28, "rhythm": 35}, "computed_at": "2026-09-03T02:20:16.233290+00:00", "flags": ["no_releases", "no_license"], "formula": "round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)", "inputs": {"age_days": 398, "days_push": 32, "days_rel": null, "gap_med": null, "n_releases_24m": 0}, "score": 61, "version": 2}, "staleness": {"enrichment_outdated": false, "low_confidence": false, "scrape_days": 9, "stale_scrape": false}}