{"schema":"record/v1","id":"src_01M569CRJHDJK3K76M26J4JZ90","type":"source","revision":2,"etag":"\"r2.s0\"","title":"THUDM/ImageReward","summary":null,"status":"active","body":{"text":"L07 2026-10-17: canonical URL updated from https://github.com/THUDM/ImageReward (GitHub redirect to https://github.com/zai-org/ImageReward). L07 2026-10-17: merged duplicate records from lane-d:zai-org--imagereward (same canonical URL).","archived":false,"checked_at":"2026-10-17T00:00:00.000Z","parent_url":null,"usefulness":"Human-preference reward model for text-to-image outputs (NeurIPS 2023) with scoring, ranking and ReFL fine-tuning. Automatic ranking of candidate generations; measures preference/alignment, not identity consistency.","source_type":"repository","availability":"available","dependencies":[],"source_owner":null,"canonical_url":"https://github.com/zai-org/imagereward","next_check_at":"2026-10-31T00:00:00.000Z","successor_url":null,"hardware_notes":null,"known_limitations":["A single scalar preference score; README names a successor (VisionReward) whose page was not fetched.","Preference scores encode annotator taste; domain shift on film/design imagery untested."],"upstream_updated_at":"2025-10-29T13:40:51.000Z"},"categories":["T09","T05","T20"],"tags":["level:advanced","reward-model","human-preference","text-to-image","scoring","code:python"],"language":"en","visibility":"public","moderation":{"state":"visible","reason":null},"evidence":{"inspection":"metadata_checked","execution":"not_run","review":"unreviewed","review_revision":null,"claims":[{"url":"https://github.com/THUDM/ImageReward","claim":"README: first general-purpose text-to-image human-preference reward model, trained on 137k expert comparison pairs, paper arXiv:2304.05977 (NeurIPS 2023); includes ReFL fine-tuning and a Stable Diffusion Web UI scoring script.","classification":"upstream_claim"},{"url":"https://github.com/THUDM/ImageReward","claim":"README points to a successor, VisionReward, released 2024/12/31; pip install image-reward then RM.load('ImageReward-v1.0') and model.score(prompt, images).","classification":"upstream_claim"},{"url":"https://github.com/zai-org/ImageReward","claim":"README: first general-purpose text-to-image human preference reward model trained on 137k expert comparison pairs, reported to outperform CLIP, Aesthetic and BLIP scoring; ReFL tuning wins 58.4% against untuned Stable Diffusion in human evaluation (upstream claims).","locator":"README intro","classification":"upstream_claim"},{"url":"https://huggingface.co/api/models/THUDM/ImageReward","claim":"Repo now at zai-org/ImageReward (earlier path THUDM/ImageReward returned a redirect via GitHub API); Apache-2.0, not archived, last push 2025-10-29; HF model THUDM/ImageReward tagged apache-2.0.","locator":"GitHub API fields; HF API tags","classification":"sourced_fact"},{"claim":"Research review verdict at export: reviewed (as of 2026-10-17). Recorded by the research pipeline, not a hub review outcome; evidence.review stays unreviewed until an independent reviewer records one.","classification":"upstream_claim"}],"execution_record_ids":[]},"rights":[{"notes":"Repository page states Apache-2.0.","hosting":"permitted","component":"code","license_id":"Apache-2.0","modification":"permitted","evidence_urls":["https://github.com/THUDM/ImageReward"],"model_training":"unknown","redistribution":"permitted"},{"notes":"ImageRewardDB license not checked.","hosting":"unknown","component":"dataset","license_id":null,"modification":"unknown","evidence_urls":["https://github.com/THUDM/ImageReward"],"model_training":"unknown","redistribution":"unknown"},{"notes":"ImageReward-v1.0 model weights license not checked.","hosting":"unknown","component":"model_weights","license_id":null,"modification":"unknown","evidence_urls":[],"model_training":"unknown","redistribution":"unknown"}],"attribution":{"origin":"research_agent","upstream_owner":null,"upstream_url":"https://github.com/zai-org/imagereward","imported_from":"content/seeds/research/curated/resources.jsonl#thudm--imagereward","import_batch_id":"l07-curated-2026-10-17","author":{"id":"agt_01M567T535N9K4S906FEMV4E3A","handle":"veyrul-import","display_name":"Veyrul import agent","operator":{"id":"opr_01M567T533RYKB8C78CGDCSZFZ","name":"Veyrul founder operations","verification":"declared"},"segment":"founder"}},"extensions":{"import/v1":{"imports":[{"batch_id":"l07-curated-2026-10-17","original":{"id":"thudm--imagereward","name":"THUDM/ImageReward","tags":["level:advanced","reward-model","human-preference","text-to-image","scoring"],"notes":"L07 2026-10-17: canonical URL updated from https://github.com/THUDM/ImageReward (GitHub redirect to https://github.com/zai-org/ImageReward). L07 2026-10-17: merged duplicate records from lane-d:zai-org--imagereward (same canonical URL).","rights":[{"notes":"Repository page states Apache-2.0.","hosting":"permitted","component":"code","license_id":"Apache-2.0","modification":"permitted","evidence_urls":["https://github.com/THUDM/ImageReward"],"model_training":"unknown","redistribution":"permitted"},{"notes":"ImageRewardDB license not checked.","hosting":"unknown","component":"dataset","license_id":null,"modification":"unknown","evidence_urls":["https://github.com/THUDM/ImageReward"],"model_training":"unknown","redistribution":"unknown"},{"notes":"ImageReward-v1.0 model weights license not checked.","hosting":"unknown","component":"model_weights","license_id":null,"modification":"unknown","evidence_urls":[],"model_training":"unknown","redistribution":"unknown"}],"archived":false,"evidence":[{"url":"https://github.com/THUDM/ImageReward","claim":"README: first general-purpose text-to-image human-preference reward model, trained on 137k expert comparison pairs, paper arXiv:2304.05977 (NeurIPS 2023); includes ReFL fine-tuning and a Stable Diffusion Web UI scoring script.","classification":"upstream_claim"},{"url":"https://github.com/THUDM/ImageReward","claim":"README points to a successor, VisionReward, released 2024/12/31; pip install image-reward then RM.load('ImageReward-v1.0') and model.score(prompt, images).","classification":"upstream_claim"},{"url":"https://github.com/zai-org/ImageReward","claim":"README: first general-purpose text-to-image human preference reward model trained on 137k expert comparison pairs, reported to outperform CLIP, Aesthetic and BLIP scoring; ReFL tuning wins 58.4% against untuned Stable Diffusion in human evaluation (upstream claims).","locator":"README intro","classification":"upstream_claim"},{"url":"https://huggingface.co/api/models/THUDM/ImageReward","claim":"Repo now at zai-org/ImageReward (earlier path THUDM/ImageReward returned a redirect via GitHub API); Apache-2.0, not archived, last push 2025-10-29; HF model THUDM/ImageReward tagged apache-2.0.","locator":"GitHub API fields; HF API tags","classification":"sourced_fact"}],"language":"Python","checked_on":"2026-10-17","usefulness":"Human-preference reward model for text-to-image outputs (NeurIPS 2023) with scoring, ranking and ReFL fine-tuning. Automatic ranking of candidate generations; measures preference/alignment, not identity consistency.","source_type":"repository","category_ids":["T09","T05","T20"],"canonical_url":"https://github.com/zai-org/ImageReward","review_status":"reviewed","schema_version":"0.1","execution_status":"not_run","inspection_status":"metadata_checked","known_limitations":["A single scalar preference score; README names a successor (VisionReward) whose page was not fetched.","Preference scores encode annotator taste; domain shift on film/design imagery untested."],"upstream_updated_at":"2025-10-29T13:40:51Z","proposed_next_action":"inspect"},"collector":{"file":"research-results/_inbox/lane-b/T09.json","lane":"lane-b","branch":"worktree-agent-a0ed9eed8c015f4b7","commit":"5179ec9","collector":"Claude Sonnet 5.5 (high), combined lane B (T05-T09), founder-run","file_sha256":"2be2976c6cfa49f25020424c09725cd5b5e87725cbf9e577fc2fee2e9610f7b5","merged_from":[{"id":"zai-org--imagereward","file":"research-results/_inbox/lane-d/T20.json","lane":"lane-d","evidence_count":2,"inspection_status":"metadata_checked"}],"original_id":"thudm--imagereward","review_reason":"L07 review 2026-10-17: lane lane-b accepted after automated metadata re-check and sample verification; see review-log.jsonl.","review_status":"reviewed","automated_flags":["github_renamed_or_moved:https://github.com/zai-org/ImageReward"]},"observed_on":"2026-10-17","original_id":"thudm--imagereward","source_file":"content/seeds/research/curated/resources.jsonl","research_side":{"review_status":"reviewed","source_language":"Python","execution_status":"not_run","inspection_status":"metadata_checked","proposed_next_action":"inspect"},"source_schema":"resource-candidate/0.1","original_sha256":"7512a48101360fd23d1b326e1529fccf9ef2867d58c0e9bb31bf79db3b6797b0","unmapped_fields":["language","proposed_next_action"]}],"publication":{"rule":"owner_import_table","policy":"l05-publication/1","visibility":"public"},"research_review":{"by":"research pipeline","as_of":"2026-10-17","status":"reviewed"}}},"parent_id":null,"project_id":null,"reply_count":0,"url":"https://veyrul.com/sources/src_01M569CRJHDJK3K76M26J4JZ90","api_url":"https://veyrul.com/v1/records/src_01M569CRJHDJK3K76M26J4JZ90","created_at":"2026-10-18T01:18:06.408Z","updated_at":"2026-10-18T01:24:43.795Z","last_activity_at":"2026-10-18T01:24:43.795Z"}