{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:X3JC64ASYRLPMMN4XIJCZCZ2Z7","short_pith_number":"pith:X3JC64AS","schema_version":"1.0","canonical_sha256":"bed22f7012c456f631bcba122c8b3acfd91202c709f2f8a95fe690a779293b1b","source":{"kind":"arxiv","id":"2506.02153","version":2},"attestation_state":"computed","paper":{"title":"Small Language Models are the Future of Agentic AI","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Small language models will replace large ones in most agentic AI applications due to better suitability and economy for specialized tasks.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Greg Heinrich, Pavlo Molchanov, Peter Belcak, Saurav Muralidharan, Shizhe Diao, Xin Dong, Yingyan Celine Lin, Yonggan Fu","submitted_at":"2025-06-02T18:35:16Z","abstract_excerpt":"Large language models (LLMs) are often praised for exhibiting near-human performance on a wide range of tasks and valued for their ability to hold a general conversation. The rise of agentic AI systems is, however, ushering in a mass of applications in which language models perform a small number of specialized tasks repetitively and with little variation.\n  Here we lay out the position that small language models (SLMs) are sufficiently powerful, inherently more suitable, and necessarily more economical for many invocations in agentic systems, and are therefore the future of agentic AI. Our ar"},"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":true},"canonical_record":{"source":{"id":"2506.02153","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-06-02T18:35:16Z","cross_cats_sorted":[],"title_canon_sha256":"0f7eb6283ded3977bbf0e5bc2f853c3034c617bbbf507a95be609e5198cdd54a","abstract_canon_sha256":"8886f67a67aecd07adfb707313cfa51995b02364b9bbee8a66f7140e6921edb5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:47.967001Z","signature_b64":"zm4ON+qMhqVFj7xDEpuXxKuzzh2t9lFFOsY67YfViXQ/WI/a059Fe2+PWqucVyzIZ7m1xDEJVBRDpWLj1K7QAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bed22f7012c456f631bcba122c8b3acfd91202c709f2f8a95fe690a779293b1b","last_reissued_at":"2026-05-17T23:38:47.966178Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:47.966178Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Small Language Models are the Future of Agentic AI","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Small language models will replace large ones in most agentic AI applications due to better suitability and economy for specialized tasks.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Greg Heinrich, Pavlo Molchanov, Peter Belcak, Saurav Muralidharan, Shizhe Diao, Xin Dong, Yingyan Celine Lin, Yonggan Fu","submitted_at":"2025-06-02T18:35:16Z","abstract_excerpt":"Large language models (LLMs) are often praised for exhibiting near-human performance on a wide range of tasks and valued for their ability to hold a general conversation. The rise of agentic AI systems is, however, ushering in a mass of applications in which language models perform a small number of specialized tasks repetitively and with little variation.\n  Here we lay out the position that small language models (SLMs) are sufficiently powerful, inherently more suitable, and necessarily more economical for many invocations in agentic systems, and are therefore the future of agentic AI. Our ar"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"SLMs are sufficiently powerful, inherently more suitable, and necessarily more economical for many invocations in agentic systems, and are therefore the future of agentic AI.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the specialized, low-variation tasks in current and near-future agentic systems do not require the full general capabilities that only large models currently provide.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Small language models are sufficiently capable, more suitable, and far more economical than large models for the repetitive tasks that dominate agentic AI systems.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Small language models will replace large ones in most agentic AI applications due to better suitability and economy for specialized tasks.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"0d7eb9264358986f9ccfbb837486d5d2217e199db4b9a472a1f337ea5d711595"},"source":{"id":"2506.02153","kind":"arxiv","version":2},"verdict":{"id":"7fab1e1b-8402-464f-a9fe-38ec28a46a70","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T11:50:04.174682Z","strongest_claim":"SLMs are sufficiently powerful, inherently more suitable, and necessarily more economical for many invocations in agentic systems, and are therefore the future of agentic AI.","one_line_summary":"Small language models are sufficiently capable, more suitable, and far more economical than large models for the repetitive tasks that dominate agentic AI systems.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the specialized, low-variation tasks in current and near-future agentic systems do not require the full general capabilities that only large models currently provide.","pith_extraction_headline":"Small language models will replace large ones in most agentic AI applications due to better suitability and economy for specialized tasks."},"references":{"count":87,"sample":[{"doi":"","year":2024,"title":"Small language models vs","work_id":"be3a86fd-412c-4a1d-ac03-6bb039283277","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Small language models vs","work_id":"451b6b21-a6ba-4be2-a46b-2866282909a5","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone","work_id":"feef9556-a016-493c-abd2-0c97a23a7ebf","ref_index":3,"cited_arxiv_id":"2404.14219","is_internal_anchor":true},{"doi":"","year":2025,"title":"The economics of ai training and inference: How deepseek broke the cost curve, February 2025","work_id":"f32ce730-def0-4ffc-b22b-20e835db4927","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Delift: Data efficient language model instruction fine tuning.arXiv preprint arXiv:2411.04425, 2024","work_id":"89e0eb30-f21e-4d31-8234-4766ca73a3e8","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":87,"snapshot_sha256":"effa9e39be785856559f97913d36c416ee3b3462e5c52921ab7ec4bb36270593","internal_anchors":11},"formal_canon":{"evidence_count":2,"snapshot_sha256":"6bb4fe5856ab8924b6beaaaee9b6cb991fc6c2964366cd853b8135f3348bf364"},"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":"2506.02153","created_at":"2026-05-17T23:38:47.966303+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.02153v2","created_at":"2026-05-17T23:38:47.966303+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.02153","created_at":"2026-05-17T23:38:47.966303+00:00"},{"alias_kind":"pith_short_12","alias_value":"X3JC64ASYRLP","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"X3JC64ASYRLPMMN4","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"X3JC64AS","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":38,"internal_anchor_count":38,"sample":[{"citing_arxiv_id":"2502.03814","citing_title":"Large Language Models for Multi-Robot Systems: A Survey","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2601.23206","citing_title":"High-quality generation of dynamic game content via small language models: A proof of concept","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2605.20706","citing_title":"Llamas on the Web: Memory-Efficient, Performance-Portable, and Multi-Precision LLM Inference with WebGPU","ref_index":5,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12991","citing_title":"Not Just RLHF: Why Alignment Alone Won't Fix Multi-Agent Sycophancy","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18535","citing_title":"Beyond Scaling: Agents Are Heading to the Edge","ref_index":5,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18067","citing_title":"PPAI: Enabling Personalized LLM Agent Interoperability for Collaborative Edge Intelligence","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18380","citing_title":"QSTRBench: a New Benchmark to Evaluate the Ability of Language Models to Reason with Qualitative Spatial and Temporal Calculi","ref_index":74,"is_internal_anchor":true},{"citing_arxiv_id":"2605.19717","citing_title":"Physics-in-the-Loop: A Hybrid Agentic Architecture for Validated CAD Engineering Design","ref_index":67,"is_internal_anchor":true},{"citing_arxiv_id":"2605.16767","citing_title":"Retrieval-Based Multi-Label Legal Annotation: Extensible, Data-Efficient and Hallucination-Free","ref_index":29,"is_internal_anchor":true},{"citing_arxiv_id":"2605.16991","citing_title":"Response-free item difficulty modelling for multiple-choice items with fine-tuned transformers: Component-wise representation and multi-task learning","ref_index":69,"is_internal_anchor":true},{"citing_arxiv_id":"2605.15206","citing_title":"AgentStop: Terminating Local AI Agents Early to Save Energy in Consumer Devices","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2508.16703","citing_title":"ShadowNPU: System and Algorithm Co-design for NPU-Centric On-Device LLM Inference","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2511.00739","citing_title":"Towards Understanding, Analyzing, and Optimizing Agentic AI Execution: A CPU-Centric Perspective","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2511.11362","citing_title":"On-Device Fine-Tuning via Backprop-Free Zeroth-Order Optimization","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2511.18258","citing_title":"Hybrid Agentic AI and Multi-Agent Systems in Smart Manufacturing","ref_index":20,"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":3,"is_internal_anchor":true},{"citing_arxiv_id":"2512.06721","citing_title":"ProAgent: Harnessing On-Demand Sensory Contexts for Proactive LLM Agent Systems in the Wild","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2512.22579","citing_title":"SANet: A Semantic-aware Agentic AI Networking Framework for Cross-layer Optimization in 6G","ref_index":35,"is_internal_anchor":true},{"citing_arxiv_id":"2601.07961","citing_title":"Language Markers of Emotion Flexibility Predict Depression and Anxiety Treatment Outcomes","ref_index":28,"is_internal_anchor":true},{"citing_arxiv_id":"2604.06173","citing_title":"Beyond Case Law: Evaluating Structure-Aware Retrieval and Safety in Statute-Centric Legal QA","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2602.11327","citing_title":"Security Threat Modeling for Emerging AI-Agent Protocols: A Comparative Analysis of MCP, A2A, Agora, and ANP","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2509.20354","citing_title":"EmbeddingGemma: Powerful and Lightweight Text Representations","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2604.04692","citing_title":"Is a Picture Worth a Thousand Words? Adaptive Multimodal Fact-Checking with Visual Evidence Necessity","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12991","citing_title":"Not Just RLHF: Why Alignment Alone Won't Fix Multi-Agent Sycophancy","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2605.03195","citing_title":"Terminus-4B: Can a Smaller Model Replace Frontier LLMs at Agentic Execution Tasks?","ref_index":17,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/X3JC64ASYRLPMMN4XIJCZCZ2Z7","json":"https://pith.science/pith/X3JC64ASYRLPMMN4XIJCZCZ2Z7.json","graph_json":"https://pith.science/api/pith-number/X3JC64ASYRLPMMN4XIJCZCZ2Z7/graph.json","events_json":"https://pith.science/api/pith-number/X3JC64ASYRLPMMN4XIJCZCZ2Z7/events.json","paper":"https://pith.science/paper/X3JC64AS"},"agent_actions":{"view_html":"https://pith.science/pith/X3JC64ASYRLPMMN4XIJCZCZ2Z7","download_json":"https://pith.science/pith/X3JC64ASYRLPMMN4XIJCZCZ2Z7.json","view_paper":"https://pith.science/paper/X3JC64AS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.02153&json=true","fetch_graph":"https://pith.science/api/pith-number/X3JC64ASYRLPMMN4XIJCZCZ2Z7/graph.json","fetch_events":"https://pith.science/api/pith-number/X3JC64ASYRLPMMN4XIJCZCZ2Z7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/X3JC64ASYRLPMMN4XIJCZCZ2Z7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/X3JC64ASYRLPMMN4XIJCZCZ2Z7/action/storage_attestation","attest_author":"https://pith.science/pith/X3JC64ASYRLPMMN4XIJCZCZ2Z7/action/author_attestation","sign_citation":"https://pith.science/pith/X3JC64ASYRLPMMN4XIJCZCZ2Z7/action/citation_signature","submit_replication":"https://pith.science/pith/X3JC64ASYRLPMMN4XIJCZCZ2Z7/action/replication_record"}},"created_at":"2026-05-17T23:38:47.966303+00:00","updated_at":"2026-05-17T23:38:47.966303+00:00"}