{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:PA3MCOFJRIH3GPJFL5NYGDNS46","short_pith_number":"pith:PA3MCOFJ","schema_version":"1.0","canonical_sha256":"7836c138a98a0fb33d255f5b830db2e78729ce4a796224c35216fc6cb736bb6e","source":{"kind":"arxiv","id":"2401.11817","version":2},"attestation_state":"computed","paper":{"title":"Hallucination is Inevitable: An Innate Limitation of Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"LLMs cannot learn all computable functions and will therefore inevitably hallucinate when used as general problem solvers.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Mohan Kankanhalli, Sanjay Jain, Ziwei Xu","submitted_at":"2024-01-22T10:26:14Z","abstract_excerpt":"Hallucination has been widely recognized to be a significant drawback for large language models (LLMs). There have been many works that attempt to reduce the extent of hallucination. These efforts have mostly been empirical so far, which cannot answer the fundamental question whether it can be completely eliminated. In this paper, we formalize the problem and show that it is impossible to eliminate hallucination in LLMs. Specifically, we define a formal world where hallucination is defined as inconsistencies between a computable LLM and a computable ground truth function. By employing results "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2401.11817","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-01-22T10:26:14Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"0a5982e2771b534222c546140a26a6e14601bed245b6e35d3b5a1fcd4f37b4cc","abstract_canon_sha256":"248c375cbb4fd5f2746f4d1b73399ec0426c9050dc6345ed0f1bc46fbd9b1bf4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:50.233404Z","signature_b64":"SJBJdktNL2/o1z+lZpdYBK+dZhteKP/trAKUrKlcwRlKyvyP54DmH3eBXE5eAqGNQF0K0I19armgFlVon0wNAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7836c138a98a0fb33d255f5b830db2e78729ce4a796224c35216fc6cb736bb6e","last_reissued_at":"2026-05-17T23:38:50.232845Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:50.232845Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hallucination is Inevitable: An Innate Limitation of Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"LLMs cannot learn all computable functions and will therefore inevitably hallucinate when used as general problem solvers.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Mohan Kankanhalli, Sanjay Jain, Ziwei Xu","submitted_at":"2024-01-22T10:26:14Z","abstract_excerpt":"Hallucination has been widely recognized to be a significant drawback for large language models (LLMs). There have been many works that attempt to reduce the extent of hallucination. These efforts have mostly been empirical so far, which cannot answer the fundamental question whether it can be completely eliminated. In this paper, we formalize the problem and show that it is impossible to eliminate hallucination in LLMs. Specifically, we define a formal world where hallucination is defined as inconsistencies between a computable LLM and a computable ground truth function. By employing results "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"we show that LLMs cannot learn all the computable functions and will therefore inevitably hallucinate if used as general problem solvers","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The formal computable world is representative enough of the real world that the impossibility result carries over directly to practical LLMs.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Hallucinations are inevitable in LLMs because they cannot learn all computable functions according to learning theory.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"LLMs cannot learn all computable functions and will therefore inevitably hallucinate when used as general problem solvers.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"106efa8013cf6b7b68909c71c131edd857bfa80a47e3f3757a9c449b3609016f"},"source":{"id":"2401.11817","kind":"arxiv","version":2},"verdict":{"id":"02ce53cb-c065-4361-b697-200316f95816","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T20:34:45.866926Z","strongest_claim":"we show that LLMs cannot learn all the computable functions and will therefore inevitably hallucinate if used as general problem solvers","one_line_summary":"Hallucinations are inevitable in LLMs because they cannot learn all computable functions according to learning theory.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The formal computable world is representative enough of the real world that the impossibility result carries over directly to practical LLMs.","pith_extraction_headline":"LLMs cannot learn all computable functions and will therefore inevitably hallucinate when used as general problem solvers."},"references":{"count":89,"sample":[{"doi":"","year":2023,"title":"Characterizing Attribution and Fluency Tradeoffs for Retrieval-Augmented Large Language Models","work_id":"3b32e10b-3701-437c-a4c0-7fbb105ee433","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2009,"title":"Computational Complexity - A Modern Approach","work_id":"888d817a-da41-4ec9-85a9-5d89879d4a1c","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1972,"title":"On the prediction of General Recursive Functions","work_id":"1aeaaba0-fc61-4e22-b67d-fda9563709e9","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Learning families of algebraic structures from informant","work_id":"9e90b2a5-4c25-429d-9d93-12491b8caa33","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Airline held liable for its chatbot giving passenger bad advice -- what this means for travellers, February 2024","work_id":"fa7b7ba5-db49-49a5-b623-9109f6c71101","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":89,"snapshot_sha256":"f80b2183ac7614a45ffbae74f7dc9fc629125114e82d2487c487cabdf250b89f","internal_anchors":10},"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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2401.11817","created_at":"2026-05-17T23:38:50.232941+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.11817v2","created_at":"2026-05-17T23:38:50.232941+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.11817","created_at":"2026-05-17T23:38:50.232941+00:00"},{"alias_kind":"pith_short_12","alias_value":"PA3MCOFJRIH3","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"PA3MCOFJRIH3GPJF","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"PA3MCOFJ","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":36,"internal_anchor_count":36,"sample":[{"citing_arxiv_id":"2605.23026","citing_title":"Opportunities and Risks of Generative AI through the Health Information Journey","ref_index":65,"is_internal_anchor":true},{"citing_arxiv_id":"2407.20240","citing_title":"Social and Ethical Risks Posed by General-Purpose LLMs for Settling Newcomers in Canada","ref_index":31,"is_internal_anchor":true},{"citing_arxiv_id":"2411.16771","citing_title":"VidHal: Benchmarking Temporal Hallucinations in Vision LLMs","ref_index":58,"is_internal_anchor":true},{"citing_arxiv_id":"2502.02871","citing_title":"Position: Multimodal Large Language Models Can Significantly Advance Scientific Reasoning","ref_index":220,"is_internal_anchor":true},{"citing_arxiv_id":"2502.07143","citing_title":"Ask Patients with Patience: Enabling LLMs for Human-Centric Medical Dialogue with Grounded Reasoning","ref_index":5,"is_internal_anchor":true},{"citing_arxiv_id":"2502.12187","citing_title":"Hallucinations are inevitable but can be made statistically negligible","ref_index":45,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18732","citing_title":"Predictable Confabulations: Factual Recall by LLMs Scales with Model Size and Topic Frequency","ref_index":25,"is_internal_anchor":true},{"citing_arxiv_id":"2506.01481","citing_title":"TSGuard: Automated User-Centric Incident Diagnosis for AI Workloads in the Cloud","ref_index":65,"is_internal_anchor":true},{"citing_arxiv_id":"2506.02546","citing_title":"To trust or not to trust: Attention-based Trust Management for LLM Multi-Agent Systems","ref_index":40,"is_internal_anchor":true},{"citing_arxiv_id":"2506.10060","citing_title":"Textual Bayes: Quantifying Prompt Uncertainty in LLM-Based Systems","ref_index":71,"is_internal_anchor":true},{"citing_arxiv_id":"2507.10722","citing_title":"Bridging Brains and Machines: A Unified Frontier in Neuroscience, Artificial Intelligence, and Neuromorphic Systems","ref_index":185,"is_internal_anchor":true},{"citing_arxiv_id":"2509.00789","citing_title":"CogDriver: Integrating Cognitive Inertia for Temporally Coherent Planning in Autonomous Driving","ref_index":44,"is_internal_anchor":true},{"citing_arxiv_id":"2509.21654","citing_title":"Limitations on Accurate, Trusted, Human-level Reasoning","ref_index":17,"is_internal_anchor":true},{"citing_arxiv_id":"2511.08877","citing_title":"Hierarchical Memorization in Large Language Models: Evidence from Citation Generation","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2511.12439","citing_title":"Multi-agent Self-triage System with Medical Flowcharts","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2512.03053","citing_title":"Mitigating hallucinations and omissions in LLMs for invertible problems: An application to hardware logic design automation","ref_index":36,"is_internal_anchor":true},{"citing_arxiv_id":"2406.20094","citing_title":"Scaling Synthetic Data Creation with 1,000,000,000 Personas","ref_index":25,"is_internal_anchor":true},{"citing_arxiv_id":"2602.20669","citing_title":"Integrating Domain-Specialized Language Models with AI Measurement Tools for Deterministic Atomic-Resolution Experimentation","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2603.13200","citing_title":"Navig-AI-tion: Navigation by Contextual AI and Spatial Audio","ref_index":25,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14517","citing_title":"Dimension-Level Intent Fidelity Evaluation for Large Language Models: Evidence from Structured Prompt Ablation","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2604.27006","citing_title":"Beyond Accuracy: LLM Variability in Evidence Screening for Software Engineering SLRs","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09278","citing_title":"EquiMem: Calibrating Shared Memory in Multi-Agent Debate via Game-Theoretic Equilibrium","ref_index":85,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10246","citing_title":"SciIntegrity-Bench: A Benchmark for Evaluating Academic Integrity in AI Scientist Systems","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2605.02994","citing_title":"Using Large Language Models as a Co-Author in Undergraduate Quantum Group Research","ref_index":28,"is_internal_anchor":true},{"citing_arxiv_id":"2604.24700","citing_title":"Green Shielding: A User-Centric Approach Towards Trustworthy AI","ref_index":5,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PA3MCOFJRIH3GPJFL5NYGDNS46","json":"https://pith.science/pith/PA3MCOFJRIH3GPJFL5NYGDNS46.json","graph_json":"https://pith.science/api/pith-number/PA3MCOFJRIH3GPJFL5NYGDNS46/graph.json","events_json":"https://pith.science/api/pith-number/PA3MCOFJRIH3GPJFL5NYGDNS46/events.json","paper":"https://pith.science/paper/PA3MCOFJ"},"agent_actions":{"view_html":"https://pith.science/pith/PA3MCOFJRIH3GPJFL5NYGDNS46","download_json":"https://pith.science/pith/PA3MCOFJRIH3GPJFL5NYGDNS46.json","view_paper":"https://pith.science/paper/PA3MCOFJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.11817&json=true","fetch_graph":"https://pith.science/api/pith-number/PA3MCOFJRIH3GPJFL5NYGDNS46/graph.json","fetch_events":"https://pith.science/api/pith-number/PA3MCOFJRIH3GPJFL5NYGDNS46/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PA3MCOFJRIH3GPJFL5NYGDNS46/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PA3MCOFJRIH3GPJFL5NYGDNS46/action/storage_attestation","attest_author":"https://pith.science/pith/PA3MCOFJRIH3GPJFL5NYGDNS46/action/author_attestation","sign_citation":"https://pith.science/pith/PA3MCOFJRIH3GPJFL5NYGDNS46/action/citation_signature","submit_replication":"https://pith.science/pith/PA3MCOFJRIH3GPJFL5NYGDNS46/action/replication_record"}},"created_at":"2026-05-17T23:38:50.232941+00:00","updated_at":"2026-05-17T23:38:50.232941+00:00"}