{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:74UNZO64LBQTVYXH53TQ6BCMWJ","short_pith_number":"pith:74UNZO64","canonical_record":{"source":{"id":"2602.09130","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-09T19:20:56Z","cross_cats_sorted":[],"title_canon_sha256":"37e31a3254c39434f8c64973d10a84c1fc62da8385b019e3b8f3e137135e0ea0","abstract_canon_sha256":"e8e643c8cf05faf8ddc3b076fdaaf3a8df66a9875148d0a653318fcbac0d25eb"},"schema_version":"1.0"},"canonical_sha256":"ff28dcbbdc58613ae2e7eee70f044cb248477f5bdd50487758d89f62a967cb24","source":{"kind":"arxiv","id":"2602.09130","version":5},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2602.09130","created_at":"2026-05-26T01:03:24Z"},{"alias_kind":"arxiv_version","alias_value":"2602.09130v5","created_at":"2026-05-26T01:03:24Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.09130","created_at":"2026-05-26T01:03:24Z"},{"alias_kind":"pith_short_12","alias_value":"74UNZO64LBQT","created_at":"2026-05-26T01:03:24Z"},{"alias_kind":"pith_short_16","alias_value":"74UNZO64LBQTVYXH","created_at":"2026-05-26T01:03:24Z"},{"alias_kind":"pith_short_8","alias_value":"74UNZO64","created_at":"2026-05-26T01:03:24Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:74UNZO64LBQTVYXH53TQ6BCMWJ","target":"record","payload":{"canonical_record":{"source":{"id":"2602.09130","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-09T19:20:56Z","cross_cats_sorted":[],"title_canon_sha256":"37e31a3254c39434f8c64973d10a84c1fc62da8385b019e3b8f3e137135e0ea0","abstract_canon_sha256":"e8e643c8cf05faf8ddc3b076fdaaf3a8df66a9875148d0a653318fcbac0d25eb"},"schema_version":"1.0"},"canonical_sha256":"ff28dcbbdc58613ae2e7eee70f044cb248477f5bdd50487758d89f62a967cb24","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:03:24.998462Z","signature_b64":"6LS13LmKim1+SADMHRLEZNLMkoecVJiXOzZ25avp7fpHJe3Em9FDTKBC6i6CF8uRPmKncTBnSdqpZQYr4dVbBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ff28dcbbdc58613ae2e7eee70f044cb248477f5bdd50487758d89f62a967cb24","last_reissued_at":"2026-05-26T01:03:24.997648Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:03:24.997648Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2602.09130","source_version":5,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-26T01:03:24Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"BlBbQ6xPf8YTFhiW4dtujQ2UDPtgl5q45Fl3P25BOElLTJuBIGTMV+Cz5ixLWnt3jz8Ig4KneeaTeuwN/6SFBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-12T07:59:44.701130Z"},"content_sha256":"f5ceb3a9ec118628a734f965544c3813304f01fa00e9ff332ce4b241dc154696","schema_version":"1.0","event_id":"sha256:f5ceb3a9ec118628a734f965544c3813304f01fa00e9ff332ce4b241dc154696"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:74UNZO64LBQTVYXH53TQ6BCMWJ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"UniComp: A Unified Evaluation of Large Language Model Compression via Pruning, Quantization and Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Compressed language models preserve factual recall but lose multi-step reasoning, multilingual ability, and instruction following.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andreas Geiger, Jonathan von Rad, Yong Cao","submitted_at":"2026-02-09T19:20:56Z","abstract_excerpt":"Model compression is increasingly essential for deploying large language models (LLMs), yet existing comparative studies largely focus on pruning and quantization evaluated primarily on knowledge-centric benchmarks. Thus, we introduce UniComp, a unified evaluation framework for comparing pruning, quantization, and knowledge distillation. UniComp evaluates compressed models along three dimensions: performance, reliability, and efficiency, using a diverse set of capability- and safety-oriented benchmarks together with a hardware-aware efficiency analysis. Through evaluation of six compression te"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Through evaluation of six compression techniques across 40 datasets, we observe (i) a consistent knowledge bias, where factual recall is largely preserved while multi-step reasoning, multilingual, and instruction-following capabilities degrade; (ii) a decoupling between performance and reliability, indicating that retained performance does not consistently imply preserved reliability; and (iii) that task-specific calibration can yield up to 50% relative improvement of reasoning performance in pruned models.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The selected benchmarks and compression techniques are representative of general behavior, and the chosen reliability metrics accurately capture real-world safety and consistency issues without hidden selection effects.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"UniComp finds that LLM compression preserves factual recall but degrades multi-step reasoning, multilingual ability, and reliability, while task-specific calibration recovers up to 50% of lost reasoning performance in pruned models.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Compressed language models preserve factual recall but lose multi-step reasoning, multilingual ability, and instruction following.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"8d33721e483c030cc30b4a4092654f81218cfce599fe4257ed8a2d7d4fb3569e"},"source":{"id":"2602.09130","kind":"arxiv","version":5},"verdict":{"id":"ad7df9ec-4402-420e-b103-4ac03a2977b8","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T05:13:32.624478Z","strongest_claim":"Through evaluation of six compression techniques across 40 datasets, we observe (i) a consistent knowledge bias, where factual recall is largely preserved while multi-step reasoning, multilingual, and instruction-following capabilities degrade; (ii) a decoupling between performance and reliability, indicating that retained performance does not consistently imply preserved reliability; and (iii) that task-specific calibration can yield up to 50% relative improvement of reasoning performance in pruned models.","one_line_summary":"UniComp finds that LLM compression preserves factual recall but degrades multi-step reasoning, multilingual ability, and reliability, while task-specific calibration recovers up to 50% of lost reasoning performance in pruned models.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The selected benchmarks and compression techniques are representative of general behavior, and the chosen reliability metrics accurately capture real-world safety and consistency issues without hidden selection effects.","pith_extraction_headline":"Compressed language models preserve factual recall but lose multi-step reasoning, multilingual ability, and instruction following."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2602.09130/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":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"ad7df9ec-4402-420e-b103-4ac03a2977b8"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-26T01:03:24Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"GJU/9U0gX5p16KWFB+6tlogZL7K3iV8Dxf1eYv9nAQ/s+t1k7NUEcTZXMuqxcNKRLlU49divqftzh8gt/PqmBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-12T07:59:44.701627Z"},"content_sha256":"964875647fdf6aa7697cd00f94715ee8e270343d68286f51b85f38f9a9b1a3b2","schema_version":"1.0","event_id":"sha256:964875647fdf6aa7697cd00f94715ee8e270343d68286f51b85f38f9a9b1a3b2"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/74UNZO64LBQTVYXH53TQ6BCMWJ/bundle.json","state_url":"https://pith.science/pith/74UNZO64LBQTVYXH53TQ6BCMWJ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/74UNZO64LBQTVYXH53TQ6BCMWJ/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-06-12T07:59:44Z","links":{"resolver":"https://pith.science/pith/74UNZO64LBQTVYXH53TQ6BCMWJ","bundle":"https://pith.science/pith/74UNZO64LBQTVYXH53TQ6BCMWJ/bundle.json","state":"https://pith.science/pith/74UNZO64LBQTVYXH53TQ6BCMWJ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/74UNZO64LBQTVYXH53TQ6BCMWJ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:74UNZO64LBQTVYXH53TQ6BCMWJ","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"e8e643c8cf05faf8ddc3b076fdaaf3a8df66a9875148d0a653318fcbac0d25eb","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-09T19:20:56Z","title_canon_sha256":"37e31a3254c39434f8c64973d10a84c1fc62da8385b019e3b8f3e137135e0ea0"},"schema_version":"1.0","source":{"id":"2602.09130","kind":"arxiv","version":5}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2602.09130","created_at":"2026-05-26T01:03:24Z"},{"alias_kind":"arxiv_version","alias_value":"2602.09130v5","created_at":"2026-05-26T01:03:24Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.09130","created_at":"2026-05-26T01:03:24Z"},{"alias_kind":"pith_short_12","alias_value":"74UNZO64LBQT","created_at":"2026-05-26T01:03:24Z"},{"alias_kind":"pith_short_16","alias_value":"74UNZO64LBQTVYXH","created_at":"2026-05-26T01:03:24Z"},{"alias_kind":"pith_short_8","alias_value":"74UNZO64","created_at":"2026-05-26T01:03:24Z"}],"graph_snapshots":[{"event_id":"sha256:964875647fdf6aa7697cd00f94715ee8e270343d68286f51b85f38f9a9b1a3b2","target":"graph","created_at":"2026-05-26T01:03:24Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Through evaluation of six compression techniques across 40 datasets, we observe (i) a consistent knowledge bias, where factual recall is largely preserved while multi-step reasoning, multilingual, and instruction-following capabilities degrade; (ii) a decoupling between performance and reliability, indicating that retained performance does not consistently imply preserved reliability; and (iii) that task-specific calibration can yield up to 50% relative improvement of reasoning performance in pruned models."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The selected benchmarks and compression techniques are representative of general behavior, and the chosen reliability metrics accurately capture real-world safety and consistency issues without hidden selection effects."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"UniComp finds that LLM compression preserves factual recall but degrades multi-step reasoning, multilingual ability, and reliability, while task-specific calibration recovers up to 50% of lost reasoning performance in pruned models."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Compressed language models preserve factual recall but lose multi-step reasoning, multilingual ability, and instruction following."}],"snapshot_sha256":"8d33721e483c030cc30b4a4092654f81218cfce599fe4257ed8a2d7d4fb3569e"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2602.09130/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Model compression is increasingly essential for deploying large language models (LLMs), yet existing comparative studies largely focus on pruning and quantization evaluated primarily on knowledge-centric benchmarks. Thus, we introduce UniComp, a unified evaluation framework for comparing pruning, quantization, and knowledge distillation. UniComp evaluates compressed models along three dimensions: performance, reliability, and efficiency, using a diverse set of capability- and safety-oriented benchmarks together with a hardware-aware efficiency analysis. Through evaluation of six compression te","authors_text":"Andreas Geiger, Jonathan von Rad, Yong Cao","cross_cats":[],"headline":"Compressed language models preserve factual recall but lose multi-step reasoning, multilingual ability, and instruction following.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-09T19:20:56Z","title":"UniComp: A Unified Evaluation of Large Language Model Compression via Pruning, Quantization and Distillation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.09130","kind":"arxiv","version":5},"verdict":{"created_at":"2026-05-16T05:13:32.624478Z","id":"ad7df9ec-4402-420e-b103-4ac03a2977b8","model_set":{"reader":"grok-4.3"},"one_line_summary":"UniComp finds that LLM compression preserves factual recall but degrades multi-step reasoning, multilingual ability, and reliability, while task-specific calibration recovers up to 50% of lost reasoning performance in pruned models.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Compressed language models preserve factual recall but lose multi-step reasoning, multilingual ability, and instruction following.","strongest_claim":"Through evaluation of six compression techniques across 40 datasets, we observe (i) a consistent knowledge bias, where factual recall is largely preserved while multi-step reasoning, multilingual, and instruction-following capabilities degrade; (ii) a decoupling between performance and reliability, indicating that retained performance does not consistently imply preserved reliability; and (iii) that task-specific calibration can yield up to 50% relative improvement of reasoning performance in pruned models.","weakest_assumption":"The selected benchmarks and compression techniques are representative of general behavior, and the chosen reliability metrics accurately capture real-world safety and consistency issues without hidden selection effects."}},"verdict_id":"ad7df9ec-4402-420e-b103-4ac03a2977b8"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:f5ceb3a9ec118628a734f965544c3813304f01fa00e9ff332ce4b241dc154696","target":"record","created_at":"2026-05-26T01:03:24Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"e8e643c8cf05faf8ddc3b076fdaaf3a8df66a9875148d0a653318fcbac0d25eb","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-09T19:20:56Z","title_canon_sha256":"37e31a3254c39434f8c64973d10a84c1fc62da8385b019e3b8f3e137135e0ea0"},"schema_version":"1.0","source":{"id":"2602.09130","kind":"arxiv","version":5}},"canonical_sha256":"ff28dcbbdc58613ae2e7eee70f044cb248477f5bdd50487758d89f62a967cb24","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"ff28dcbbdc58613ae2e7eee70f044cb248477f5bdd50487758d89f62a967cb24","first_computed_at":"2026-05-26T01:03:24.997648Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-26T01:03:24.997648Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"6LS13LmKim1+SADMHRLEZNLMkoecVJiXOzZ25avp7fpHJe3Em9FDTKBC6i6CF8uRPmKncTBnSdqpZQYr4dVbBA==","signature_status":"signed_v1","signed_at":"2026-05-26T01:03:24.998462Z","signed_message":"canonical_sha256_bytes"},"source_id":"2602.09130","source_kind":"arxiv","source_version":5}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f5ceb3a9ec118628a734f965544c3813304f01fa00e9ff332ce4b241dc154696","sha256:964875647fdf6aa7697cd00f94715ee8e270343d68286f51b85f38f9a9b1a3b2"],"state_sha256":"dabfb29fc83b9b4ae4f75ec90e955d83f45e30cd817d4cb23660554c4a3e00b7"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"CvDrcV7ru/EANarVDU1d55mAd6KazmyPlhVCCSZ1JgUNQUO4SuR/br2sJJMtF+CjsCDwJ/TCNB+63FCULEqbAA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-12T07:59:44.703834Z","bundle_sha256":"f7a2bb081d2c33a2efb698bd792860e6bfec5a15d3d788637a11363d994933b6"}}