{"paper":{"title":"Gradients with Respect to Semantics Preserving Embeddings Tell the Uncertainty of Large Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Gradients with respect to semantics-preserving embeddings quantify uncertainty in LLM free-form generation.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Mingda Li, Rundong Lv, Ting Liu, Weinan Zhang, Xinyu Li","submitted_at":"2026-05-06T08:30:40Z","abstract_excerpt":"Uncertainty quantification (UQ) is an important technique for ensuring the trustworthiness of LLMs, given their tendency to hallucinate. Existing state-of-the-art UQ approaches for free-form generation rely heavily on sampling, which incurs high computational cost and variance. In this work, we propose the first gradient-based UQ method for free-form generation, SemGrad, which is sampling-free and computationally efficient. Unlike prior gradient-based methods developed for classification tasks that operates in parameter space, we propose to consider gradients in semantic space. Our method buil"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We propose the first gradient-based UQ method for free-form generation, SemGrad, which is sampling-free and computationally efficient. ... Experiments demonstrate that both of our methods provide efficient and effective uncertainty estimates, achieving superior performance than state-of-the-art methods, particularly in settings with multiple valid responses.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"A confident LLM maintains stable output distributions under semantically equivalent input perturbations, and the Semantic Preservation Score reliably identifies the embeddings that best capture semantics for gradient computation.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SemGrad is a gradient-based uncertainty quantification technique for free-form LLM generation that operates in semantic space using a Semantic Preservation Score to select stable embeddings.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Gradients with respect to semantics-preserving embeddings quantify uncertainty in LLM free-form generation.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"b199756e70e5f6958ffc2cd25eea21c5a56bbe9be50f0ca9b102ff2cad61b665"},"source":{"id":"2605.04638","kind":"arxiv","version":2},"verdict":{"id":"1b622d72-ca05-42c5-821b-9dce87edf581","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T16:19:59.980650Z","strongest_claim":"We propose the first gradient-based UQ method for free-form generation, SemGrad, which is sampling-free and computationally efficient. ... Experiments demonstrate that both of our methods provide efficient and effective uncertainty estimates, achieving superior performance than state-of-the-art methods, particularly in settings with multiple valid responses.","one_line_summary":"SemGrad is a gradient-based uncertainty quantification technique for free-form LLM generation that operates in semantic space using a Semantic Preservation Score to select stable embeddings.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"A confident LLM maintains stable output distributions under semantically equivalent input perturbations, and the Semantic Preservation Score reliably identifies the embeddings that best capture semantics for gradient computation.","pith_extraction_headline":"Gradients with respect to semantics-preserving embeddings quantify uncertainty in LLM free-form generation."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.04638/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T11:36:28.579702Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T22:31:19.928091Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T14:17:33.483476Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"239a208e4ba727ad979e7ac9dd3bbe762f8df2bbb0b07db41e13bbe60715342d"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}