{"paper":{"title":"Bi-CoG: Bi-Consistency-Guided Self-Training for Vision-Language Models","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"Bi-CoG produces higher-quality pseudo-labels for vision-language models by checking consistency both across models and inside a single model.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Lan-Zhe Guo, Rui Zhu, Song-Lin Lv, Zi-Kang Wang","submitted_at":"2025-10-23T12:16:41Z","abstract_excerpt":"Exploiting unlabeled data through semi-supervised learning (SSL) or leveraging pre-trained models via fine-tuning are two prevailing paradigms for addressing label-scarce scenarios. Recently, growing attention has been given to combining fine-tuning of pre-trained vision-language models (VLMs) with SSL, forming the emerging paradigm of semi-supervised fine-tuning. However, existing methods often suffer from model bias and hyperparameter sensitivity, due to reliance on prediction consistency or pre-defined confidence thresholds. To address these limitations, we propose a simple yet effective pl"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Bi-CoG assigns high-quality and low-bias pseudo-labels by simultaneously exploiting inter-model and intra-model consistency, along with an error-aware dynamic pseudo-label assignment strategy, and consistently and significantly improves the performance of existing methods over 14 datasets.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that simultaneous inter-model and intra-model consistency, together with the error-aware dynamic strategy, reliably produces pseudo-labels that are both high-quality and low-bias without introducing new forms of confirmation bias or hyperparameter sensitivity, as implied by the claim that this addresses the limitations of prior consistency or threshold-based methods.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Bi-CoG improves semi-supervised fine-tuning of vision-language models by assigning higher-quality pseudo-labels through simultaneous inter-model and intra-model consistency checks combined with dynamic error-aware selection.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Bi-CoG produces higher-quality pseudo-labels for vision-language models by checking consistency both across models and inside a single model.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"af0952d0e81abc451dc05103234a78c1ed22fa1ee32fbc54fa43aecedce86449"},"source":{"id":"2510.20477","kind":"arxiv","version":3},"verdict":{"id":"1437b69b-781e-422f-b556-67a3d56d6ac4","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-18T04:55:53.717476Z","strongest_claim":"Bi-CoG assigns high-quality and low-bias pseudo-labels by simultaneously exploiting inter-model and intra-model consistency, along with an error-aware dynamic pseudo-label assignment strategy, and consistently and significantly improves the performance of existing methods over 14 datasets.","one_line_summary":"Bi-CoG improves semi-supervised fine-tuning of vision-language models by assigning higher-quality pseudo-labels through simultaneous inter-model and intra-model consistency checks combined with dynamic error-aware selection.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that simultaneous inter-model and intra-model consistency, together with the error-aware dynamic strategy, reliably produces pseudo-labels that are both high-quality and low-bias without introducing new forms of confirmation bias or hyperparameter sensitivity, as implied by the claim that this addresses the limitations of prior consistency or threshold-based methods.","pith_extraction_headline":"Bi-CoG produces higher-quality pseudo-labels for vision-language models by checking consistency both across models and inside a single model."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2510.20477/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":1,"snapshot_sha256":"b6e8e26d98dca957144a0d4a831dc88a15e618dd0030369dddf3665f8a92456f"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}