{"paper":{"title":"Brain-CLIPLM: Semantic Compression for EEG-to-Text Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"EEG signals encode compressed semantic anchors rather than full sentence structure, making direct reconstruction overparameterized.","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.CL","authors_text":"Gang Pan, Huiyuan Tian, Jianyu Zhang, Shijian Li, Xiaoli Yang, Yurui Li","submitted_at":"2026-03-23T11:45:51Z","abstract_excerpt":"Decoding natural language from non-invasive electroencephalography (EEG) remains constrained by low signal-to-noise ratio and limited information bandwidth. This raises a central question: can sentence-level language be reliably recovered from such signals? Under realistic information constraints, this direct-recovery assumption may be too strong. We introduce a semantic compression hypothesis: non-invasive EEG may preserve recoverable semantic anchors rather than the full lexical--syntactic form of a sentence. From this perspective, direct sentence reconstruction is overly fine-grained relati"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"EEG signals encode a compressed set of semantic anchors rather than full linguistic structure, making direct sentence reconstruction overparameterized relative to EEG information capacity; Brain-CLIPLM achieves 67.55% top-5 and 85.00% top-25 sentence retrieval accuracy on the Zurich Cognitive Language Processing Corpus via granularity matching.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The semantic compression hypothesis that EEG encodes only compressed semantic anchors under realistic information constraints, and that the two-stage contrastive-plus-LLM framework aligns decoding complexity with neural capacity without losing essential information.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Brain-CLIPLM recovers compressed semantic anchors from EEG via contrastive learning and uses retrieval-grounded LLM reasoning to achieve 67.55% top-5 and 85% top-25 sentence retrieval accuracy, supporting the view that EEG encodes semantic content rather than full linguistic structure.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"EEG signals encode compressed semantic anchors rather than full sentence structure, making direct reconstruction overparameterized.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"dbcb04812db2c2dc2782d6b86473f00ba1ab8c870c973453229cfaa9280afeb0"},"source":{"id":"2604.16370","kind":"arxiv","version":2},"verdict":{"id":"da1ea1f6-25f9-4ec9-8e6f-6c63c1aa3eb1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T01:03:18.292196Z","strongest_claim":"EEG signals encode a compressed set of semantic anchors rather than full linguistic structure, making direct sentence reconstruction overparameterized relative to EEG information capacity; Brain-CLIPLM achieves 67.55% top-5 and 85.00% top-25 sentence retrieval accuracy on the Zurich Cognitive Language Processing Corpus via granularity matching.","one_line_summary":"Brain-CLIPLM recovers compressed semantic anchors from EEG via contrastive learning and uses retrieval-grounded LLM reasoning to achieve 67.55% top-5 and 85% top-25 sentence retrieval accuracy, supporting the view that EEG encodes semantic content rather than full linguistic structure.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The semantic compression hypothesis that EEG encodes only compressed semantic anchors under realistic information constraints, and that the two-stage contrastive-plus-LLM framework aligns decoding complexity with neural capacity without losing essential information.","pith_extraction_headline":"EEG signals encode compressed semantic anchors rather than full sentence structure, making direct reconstruction overparameterized."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.16370/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":2,"snapshot_sha256":"52f971d395ecbe123c5d02a67f1918eee5e4f09ae25c7836697750525cfe2132"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}