{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2014:KZQQ3R6MQZLMDEPSWUDJX22IMT","short_pith_number":"pith:KZQQ3R6M","schema_version":"1.0","canonical_sha256":"56610dc7cc8656c191f2b5069beb4864c29131d16aca960b08833884d43ee862","source":{"kind":"arxiv","id":"1412.6550","version":4},"attestation_state":"computed","paper":{"title":"FitNets: Hints for Thin Deep Nets","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A deeper but much thinner student network can outperform its larger teacher by using intermediate layer hints during training.","cross_cats":["cs.NE"],"primary_cat":"cs.LG","authors_text":"Adriana Romero, Antoine Chassang, Carlo Gatta, Nicolas Ballas, Samira Ebrahimi Kahou, Yoshua Bengio","submitted_at":"2014-12-19T22:40:51Z","abstract_excerpt":"While depth tends to improve network performances, it also makes gradient-based training more difficult since deeper networks tend to be more non-linear. The recently proposed knowledge distillation approach is aimed at obtaining small and fast-to-execute models, and it has shown that a student network could imitate the soft output of a larger teacher network or ensemble of networks. In this paper, we extend this idea to allow the training of a student that is deeper and thinner than the teacher, using not only the outputs but also the intermediate representations learned by the teacher as hin"},"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":false,"formal_links_present":true},"canonical_record":{"source":{"id":"1412.6550","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2014-12-19T22:40:51Z","cross_cats_sorted":["cs.NE"],"title_canon_sha256":"c3bc2e609ab2b30436e11b3b9d743c43e32d29b3d00eab409106a628206b534c","abstract_canon_sha256":"b5c36af2ae418d3cbb6cb525c0fd90318369ba1593e5468e6d797db8d6eb1e00"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T02:48:56.381790Z","signature_b64":"0WmxrovYUcrlDxZoba52pPr+FPWfUAaaCsMShkKoJPscJ6kH978XhFhQv7pFyEFw0hqi5UDYFR+7lTLXpARYAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"56610dc7cc8656c191f2b5069beb4864c29131d16aca960b08833884d43ee862","last_reissued_at":"2026-05-18T02:48:56.381163Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T02:48:56.381163Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FitNets: Hints for Thin Deep Nets","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A deeper but much thinner student network can outperform its larger teacher by using intermediate layer hints during training.","cross_cats":["cs.NE"],"primary_cat":"cs.LG","authors_text":"Adriana Romero, Antoine Chassang, Carlo Gatta, Nicolas Ballas, Samira Ebrahimi Kahou, Yoshua Bengio","submitted_at":"2014-12-19T22:40:51Z","abstract_excerpt":"While depth tends to improve network performances, it also makes gradient-based training more difficult since deeper networks tend to be more non-linear. The recently proposed knowledge distillation approach is aimed at obtaining small and fast-to-execute models, and it has shown that a student network could imitate the soft output of a larger teacher network or ensemble of networks. In this paper, we extend this idea to allow the training of a student that is deeper and thinner than the teacher, using not only the outputs but also the intermediate representations learned by the teacher as hin"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"For example, on CIFAR-10, a deep student network with almost 10.4 times less parameters outperforms a larger, state-of-the-art teacher network.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the added mapping parameters can reliably transfer useful intermediate knowledge from teacher to a much smaller student layer without the extra capacity causing overfitting or unstable training.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"FitNets trains deeper thinner students using teacher intermediate representations as hints plus mapping parameters, yielding a CIFAR-10 student with 10.4x fewer parameters that beats the teacher.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A deeper but much thinner student network can outperform its larger teacher by using intermediate layer hints during training.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"60684e41e36e5149e25cc689a176be470ddf8e46d03f9ec70cdbd45de6cbee81"},"source":{"id":"1412.6550","kind":"arxiv","version":4},"verdict":{"id":"51b0bf8b-5985-4814-b3a2-fdf7890eeb68","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T01:41:27.991454Z","strongest_claim":"For example, on CIFAR-10, a deep student network with almost 10.4 times less parameters outperforms a larger, state-of-the-art teacher network.","one_line_summary":"FitNets trains deeper thinner students using teacher intermediate representations as hints plus mapping parameters, yielding a CIFAR-10 student with 10.4x fewer parameters that beats the teacher.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the added mapping parameters can reliably transfer useful intermediate knowledge from teacher to a much smaller student layer without the extra capacity causing overfitting or unstable training.","pith_extraction_headline":"A deeper but much thinner student network can outperform its larger teacher by using intermediate layer hints during training."},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"d9c0ff73509e7ea2479d29d13139e226c273d36054b129fa0e2ea1185e6392fc"},"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":"1412.6550","created_at":"2026-05-18T02:48:56.381242+00:00"},{"alias_kind":"arxiv_version","alias_value":"1412.6550v4","created_at":"2026-05-18T02:48:56.381242+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1412.6550","created_at":"2026-05-18T02:48:56.381242+00:00"},{"alias_kind":"pith_short_12","alias_value":"KZQQ3R6MQZLM","created_at":"2026-05-18T12:28:35.611951+00:00"},{"alias_kind":"pith_short_16","alias_value":"KZQQ3R6MQZLMDEPS","created_at":"2026-05-18T12:28:35.611951+00:00"},{"alias_kind":"pith_short_8","alias_value":"KZQQ3R6M","created_at":"2026-05-18T12:28:35.611951+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":38,"internal_anchor_count":38,"sample":[{"citing_arxiv_id":"1906.08467","citing_title":"GAN-Knowledge Distillation for one-stage Object Detection","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"1907.00274","citing_title":"NetTailor: Tuning the Architecture, Not Just the Weights","ref_index":53,"is_internal_anchor":true},{"citing_arxiv_id":"1907.02226","citing_title":"Graph-based Knowledge Distillation by Multi-head Attention Network","ref_index":23,"is_internal_anchor":true},{"citing_arxiv_id":"1907.10804","citing_title":"Co-Evolutionary Compression for Unpaired Image Translation","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2112.11447","citing_title":"Multi-Modality Distillation via Learning the teacher's modality-level Gram Matrix","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2410.10247","citing_title":"LPT: Less-overfitting Prompt Tuning for Vision-Language Model","ref_index":31,"is_internal_anchor":true},{"citing_arxiv_id":"2605.21699","citing_title":"X-Token: Projection-Guided Cross-Tokenizer Knowledge Distillation","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2506.13127","citing_title":"Leveraging Local and Global Knowledge Integration with Time-Frequency Calibrated Distillation for Speech Enhancement","ref_index":33,"is_internal_anchor":true},{"citing_arxiv_id":"2602.04381","citing_title":"Enabling Real-Time Colonoscopic Polyp Segmentation on Commodity CPUs via Ultra-Lightweight Architecture","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2603.29167","citing_title":"JDCNet: Confidence-Gated Privileged-Modality Distillation for Cost-Preserving X-ray Inference","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2605.20681","citing_title":"Scale-Calibrated Median-of-Means for Robust Distributed Principal Component Analysis","ref_index":299,"is_internal_anchor":true},{"citing_arxiv_id":"2605.15689","citing_title":"How to Choose Your Teacher for Fine Grained Image Recognition","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2510.19239","citing_title":"TinyUSFM: Towards Compact and Efficient Ultrasound Foundation Models","ref_index":19,"is_internal_anchor":true},{"citing_arxiv_id":"2101.02388","citing_title":"Knowledge Distillation in Iterative Generative Models for Improved Sampling Speed","ref_index":30,"is_internal_anchor":true},{"citing_arxiv_id":"2512.01390","citing_title":"FRAMER: Frequency-Aligned Self-Distillation with Adaptive Modulation Leveraging Diffusion Priors for Real-World Image Super-Resolution","ref_index":36,"is_internal_anchor":true},{"citing_arxiv_id":"2507.11539","citing_title":"Streaming 4D Visual Geometry Transformer","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14071","citing_title":"Distribution Corrected Offline Data Distillation for Large Language Models","ref_index":31,"is_internal_anchor":true},{"citing_arxiv_id":"2303.17760","citing_title":"CAMEL: Communicative Agents for \"Mind\" Exploration of Large Language Model Society","ref_index":95,"is_internal_anchor":true},{"citing_arxiv_id":"1605.07146","citing_title":"Wide Residual Networks","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"2604.26857","citing_title":"Edge AI for Automotive Vulnerable Road User Safety: Deployable Detection via Knowledge Distillation","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2604.26255","citing_title":"GaitKD: A Universal Decoupled Distillation Framework for Efficient Gait Recognition","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2605.08857","citing_title":"RareCP: Regime-Aware Retrieval for Efficient Conformal Prediction","ref_index":46,"is_internal_anchor":true},{"citing_arxiv_id":"2604.26095","citing_title":"Distill-Belief: Closed-Loop Inverse Source Localization and Characterization in Physical Fields","ref_index":64,"is_internal_anchor":true},{"citing_arxiv_id":"2604.25795","citing_title":"Improving Diversity in Black-box Few-shot Knowledge Distillation","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2605.04877","citing_title":"To Fuse or to Drop? Dual-Path Learning for Resolving Modality Conflicts in Multimodal Emotion Recognition","ref_index":22,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KZQQ3R6MQZLMDEPSWUDJX22IMT","json":"https://pith.science/pith/KZQQ3R6MQZLMDEPSWUDJX22IMT.json","graph_json":"https://pith.science/api/pith-number/KZQQ3R6MQZLMDEPSWUDJX22IMT/graph.json","events_json":"https://pith.science/api/pith-number/KZQQ3R6MQZLMDEPSWUDJX22IMT/events.json","paper":"https://pith.science/paper/KZQQ3R6M"},"agent_actions":{"view_html":"https://pith.science/pith/KZQQ3R6MQZLMDEPSWUDJX22IMT","download_json":"https://pith.science/pith/KZQQ3R6MQZLMDEPSWUDJX22IMT.json","view_paper":"https://pith.science/paper/KZQQ3R6M","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1412.6550&json=true","fetch_graph":"https://pith.science/api/pith-number/KZQQ3R6MQZLMDEPSWUDJX22IMT/graph.json","fetch_events":"https://pith.science/api/pith-number/KZQQ3R6MQZLMDEPSWUDJX22IMT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KZQQ3R6MQZLMDEPSWUDJX22IMT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KZQQ3R6MQZLMDEPSWUDJX22IMT/action/storage_attestation","attest_author":"https://pith.science/pith/KZQQ3R6MQZLMDEPSWUDJX22IMT/action/author_attestation","sign_citation":"https://pith.science/pith/KZQQ3R6MQZLMDEPSWUDJX22IMT/action/citation_signature","submit_replication":"https://pith.science/pith/KZQQ3R6MQZLMDEPSWUDJX22IMT/action/replication_record"}},"created_at":"2026-05-18T02:48:56.381242+00:00","updated_at":"2026-05-18T02:48:56.381242+00:00"}