{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:E4XUT7DCORXNTYCWPBSMFOIO6R","short_pith_number":"pith:E4XUT7DC","canonical_record":{"source":{"id":"2604.06893","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-08T09:48:31Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"07683473bfea90c02f26fa281ab5069e6c583591810975e203374feafd9e0a4f","abstract_canon_sha256":"5586c1b3fa59b57633a4bdb5f4448cfa91bb0f12ba54c89eda943f7420264279"},"schema_version":"1.0"},"canonical_sha256":"272f49fc62746ed9e0567864c2b90ef47d3aa4298c4cd439809b1e85a3890bd3","source":{"kind":"arxiv","id":"2604.06893","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.06893","created_at":"2026-06-09T02:08:41Z"},{"alias_kind":"arxiv_version","alias_value":"2604.06893v3","created_at":"2026-06-09T02:08:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.06893","created_at":"2026-06-09T02:08:41Z"},{"alias_kind":"pith_short_12","alias_value":"E4XUT7DCORXN","created_at":"2026-06-09T02:08:41Z"},{"alias_kind":"pith_short_16","alias_value":"E4XUT7DCORXNTYCW","created_at":"2026-06-09T02:08:41Z"},{"alias_kind":"pith_short_8","alias_value":"E4XUT7DC","created_at":"2026-06-09T02:08:41Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:E4XUT7DCORXNTYCWPBSMFOIO6R","target":"record","payload":{"canonical_record":{"source":{"id":"2604.06893","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-08T09:48:31Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"07683473bfea90c02f26fa281ab5069e6c583591810975e203374feafd9e0a4f","abstract_canon_sha256":"5586c1b3fa59b57633a4bdb5f4448cfa91bb0f12ba54c89eda943f7420264279"},"schema_version":"1.0"},"canonical_sha256":"272f49fc62746ed9e0567864c2b90ef47d3aa4298c4cd439809b1e85a3890bd3","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T02:08:41.780571Z","signature_b64":"mvpNLP3iA7oa5xcbpEjv5nivicMGULvEziwMu3vLOomdcmwWaGk6ZDl9kjzBLATiipY3KpRZEehUiKIpLWCiCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"272f49fc62746ed9e0567864c2b90ef47d3aa4298c4cd439809b1e85a3890bd3","last_reissued_at":"2026-06-09T02:08:41.779309Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T02:08:41.779309Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.06893","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-09T02:08:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"r/8+YxTH3IbA60B6421Xg7t3XRphNCIdyqVRBLW3BTqPZWj1Dj9p17NhA+dXPrgegoCmvw45mKQuca8ZlzTeDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-11T22:38:00.559470Z"},"content_sha256":"09c3bc42af185a679e13e3b1cc2c2cca6c693d9924ef8ce033ecfd3ada562de8","schema_version":"1.0","event_id":"sha256:09c3bc42af185a679e13e3b1cc2c2cca6c693d9924ef8ce033ecfd3ada562de8"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:E4XUT7DCORXNTYCWPBSMFOIO6R","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Energy-Regularized Spatial Masking: A Novel Approach to Enhancing Robustness and Interpretability in Vision Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Embedding a differentiable energy minimization layer inside convolutional networks lets them autonomously select sparse, coherent spatial features for improved robustness and interpretability.","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Bilal Faye, Djamel Bouchaffra, Hanane Azzag, Mustapha Lebbah, Nadjib Lazaar, Tom Devynck","submitted_at":"2026-04-08T09:48:31Z","abstract_excerpt":"Deep convolutional neural networks achieve remarkable performance by exhaustively processing dense spatial feature maps, yet this brute-force strategy introduces significant computational redundancy and encourages reliance on spurious background correlations. As a result, modern vision models remain brittle and difficult to interpret. We propose Energy-Regularized Spatial Masking (ERSM), a novel framework that reformulates feature selection as a differentiable energy minimization problem. By embedding a lightweight Energy-Mask Layer inside standard convolutional backbones, each visual token is"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We validate ERSM on convolutional architectures and demonstrate that it produces emergent sparsity, improved robustness to structured occlusion, and highly interpretable spatial masks, while preserving classification accuracy. Furthermore, we show that the learned energy ranking significantly outperforms magnitude-based pruning in deletion-based robustness tests.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the proposed unary importance cost and pairwise spatial coherence penalty can be combined into a differentiable energy function whose minimization inside standard backbones yields stable training and semantically meaningful masks without additional supervision or post-hoc tuning.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ERSM reformulates spatial feature selection in vision models as energy minimization with unary importance and pairwise coherence terms, producing emergent sparsity and better occlusion robustness.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Embedding a differentiable energy minimization layer inside convolutional networks lets them autonomously select sparse, coherent spatial features for improved robustness and interpretability.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"7d99a7b0d326b084fa8d2a22b53ff148fda93ecf635bf4836ecb851dfbca1b6d"},"source":{"id":"2604.06893","kind":"arxiv","version":3},"verdict":{"id":"c73b233c-5ed2-4ea3-8cda-f0a5289b8e2d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T18:47:43.923821Z","strongest_claim":"We validate ERSM on convolutional architectures and demonstrate that it produces emergent sparsity, improved robustness to structured occlusion, and highly interpretable spatial masks, while preserving classification accuracy. Furthermore, we show that the learned energy ranking significantly outperforms magnitude-based pruning in deletion-based robustness tests.","one_line_summary":"ERSM reformulates spatial feature selection in vision models as energy minimization with unary importance and pairwise coherence terms, producing emergent sparsity and better occlusion robustness.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the proposed unary importance cost and pairwise spatial coherence penalty can be combined into a differentiable energy function whose minimization inside standard backbones yields stable training and semantically meaningful masks without additional supervision or post-hoc tuning.","pith_extraction_headline":"Embedding a differentiable energy minimization layer inside convolutional networks lets them autonomously select sparse, coherent spatial features for improved robustness and interpretability."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.06893/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":2,"snapshot_sha256":"f2e9f1d1cf6ed930232227d40c6a5346436397df948ddbc07e86d78f73650597"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"c73b233c-5ed2-4ea3-8cda-f0a5289b8e2d"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-09T02:08:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"zUgcm++9nQhiawKNOF3F2vPnf+ws6d+3KeLXda1P38airsN63H0Rce3ylhKqevqSh6EVoD3rJhzLZMKka1onDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-11T22:38:00.560435Z"},"content_sha256":"9bad3b0ee6d9d5decf7c4f9c27bc5047da9b32e6db042e201809a4f7c66992c1","schema_version":"1.0","event_id":"sha256:9bad3b0ee6d9d5decf7c4f9c27bc5047da9b32e6db042e201809a4f7c66992c1"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/E4XUT7DCORXNTYCWPBSMFOIO6R/bundle.json","state_url":"https://pith.science/pith/E4XUT7DCORXNTYCWPBSMFOIO6R/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/E4XUT7DCORXNTYCWPBSMFOIO6R/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":"2026-06-11T22:38:00Z","links":{"resolver":"https://pith.science/pith/E4XUT7DCORXNTYCWPBSMFOIO6R","bundle":"https://pith.science/pith/E4XUT7DCORXNTYCWPBSMFOIO6R/bundle.json","state":"https://pith.science/pith/E4XUT7DCORXNTYCWPBSMFOIO6R/state.json","well_known_bundle":"https://pith.science/.well-known/pith/E4XUT7DCORXNTYCWPBSMFOIO6R/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:E4XUT7DCORXNTYCWPBSMFOIO6R","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":"5586c1b3fa59b57633a4bdb5f4448cfa91bb0f12ba54c89eda943f7420264279","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-08T09:48:31Z","title_canon_sha256":"07683473bfea90c02f26fa281ab5069e6c583591810975e203374feafd9e0a4f"},"schema_version":"1.0","source":{"id":"2604.06893","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.06893","created_at":"2026-06-09T02:08:41Z"},{"alias_kind":"arxiv_version","alias_value":"2604.06893v3","created_at":"2026-06-09T02:08:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.06893","created_at":"2026-06-09T02:08:41Z"},{"alias_kind":"pith_short_12","alias_value":"E4XUT7DCORXN","created_at":"2026-06-09T02:08:41Z"},{"alias_kind":"pith_short_16","alias_value":"E4XUT7DCORXNTYCW","created_at":"2026-06-09T02:08:41Z"},{"alias_kind":"pith_short_8","alias_value":"E4XUT7DC","created_at":"2026-06-09T02:08:41Z"}],"graph_snapshots":[{"event_id":"sha256:9bad3b0ee6d9d5decf7c4f9c27bc5047da9b32e6db042e201809a4f7c66992c1","target":"graph","created_at":"2026-06-09T02:08:41Z","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":"We validate ERSM on convolutional architectures and demonstrate that it produces emergent sparsity, improved robustness to structured occlusion, and highly interpretable spatial masks, while preserving classification accuracy. Furthermore, we show that the learned energy ranking significantly outperforms magnitude-based pruning in deletion-based robustness tests."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the proposed unary importance cost and pairwise spatial coherence penalty can be combined into a differentiable energy function whose minimization inside standard backbones yields stable training and semantically meaningful masks without additional supervision or post-hoc tuning."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"ERSM reformulates spatial feature selection in vision models as energy minimization with unary importance and pairwise coherence terms, producing emergent sparsity and better occlusion robustness."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Embedding a differentiable energy minimization layer inside convolutional networks lets them autonomously select sparse, coherent spatial features for improved robustness and interpretability."}],"snapshot_sha256":"7d99a7b0d326b084fa8d2a22b53ff148fda93ecf635bf4836ecb851dfbca1b6d"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"f2e9f1d1cf6ed930232227d40c6a5346436397df948ddbc07e86d78f73650597"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.06893/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Deep convolutional neural networks achieve remarkable performance by exhaustively processing dense spatial feature maps, yet this brute-force strategy introduces significant computational redundancy and encourages reliance on spurious background correlations. As a result, modern vision models remain brittle and difficult to interpret. We propose Energy-Regularized Spatial Masking (ERSM), a novel framework that reformulates feature selection as a differentiable energy minimization problem. By embedding a lightweight Energy-Mask Layer inside standard convolutional backbones, each visual token is","authors_text":"Bilal Faye, Djamel Bouchaffra, Hanane Azzag, Mustapha Lebbah, Nadjib Lazaar, Tom Devynck","cross_cats":["cs.LG"],"headline":"Embedding a differentiable energy minimization layer inside convolutional networks lets them autonomously select sparse, coherent spatial features for improved robustness and interpretability.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-08T09:48:31Z","title":"Energy-Regularized Spatial Masking: A Novel Approach to Enhancing Robustness and Interpretability in Vision Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.06893","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-10T18:47:43.923821Z","id":"c73b233c-5ed2-4ea3-8cda-f0a5289b8e2d","model_set":{"reader":"grok-4.3"},"one_line_summary":"ERSM reformulates spatial feature selection in vision models as energy minimization with unary importance and pairwise coherence terms, producing emergent sparsity and better occlusion robustness.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Embedding a differentiable energy minimization layer inside convolutional networks lets them autonomously select sparse, coherent spatial features for improved robustness and interpretability.","strongest_claim":"We validate ERSM on convolutional architectures and demonstrate that it produces emergent sparsity, improved robustness to structured occlusion, and highly interpretable spatial masks, while preserving classification accuracy. Furthermore, we show that the learned energy ranking significantly outperforms magnitude-based pruning in deletion-based robustness tests.","weakest_assumption":"That the proposed unary importance cost and pairwise spatial coherence penalty can be combined into a differentiable energy function whose minimization inside standard backbones yields stable training and semantically meaningful masks without additional supervision or post-hoc tuning."}},"verdict_id":"c73b233c-5ed2-4ea3-8cda-f0a5289b8e2d"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:09c3bc42af185a679e13e3b1cc2c2cca6c693d9924ef8ce033ecfd3ada562de8","target":"record","created_at":"2026-06-09T02:08:41Z","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":"5586c1b3fa59b57633a4bdb5f4448cfa91bb0f12ba54c89eda943f7420264279","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-08T09:48:31Z","title_canon_sha256":"07683473bfea90c02f26fa281ab5069e6c583591810975e203374feafd9e0a4f"},"schema_version":"1.0","source":{"id":"2604.06893","kind":"arxiv","version":3}},"canonical_sha256":"272f49fc62746ed9e0567864c2b90ef47d3aa4298c4cd439809b1e85a3890bd3","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"272f49fc62746ed9e0567864c2b90ef47d3aa4298c4cd439809b1e85a3890bd3","first_computed_at":"2026-06-09T02:08:41.779309Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-09T02:08:41.779309Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"mvpNLP3iA7oa5xcbpEjv5nivicMGULvEziwMu3vLOomdcmwWaGk6ZDl9kjzBLATiipY3KpRZEehUiKIpLWCiCA==","signature_status":"signed_v1","signed_at":"2026-06-09T02:08:41.780571Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.06893","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:09c3bc42af185a679e13e3b1cc2c2cca6c693d9924ef8ce033ecfd3ada562de8","sha256:9bad3b0ee6d9d5decf7c4f9c27bc5047da9b32e6db042e201809a4f7c66992c1"],"state_sha256":"e0cbe3750d6559d737853769c5e23aa78026242e00096a76ab24bf47b463c0bc"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"V8WHwl0S7D9qwtySf9gg576O86MMW4u7FTjzWBDaqxHiw5Tu/EY5isRBfbHvnun/Wnsj3kjO260qzPNbfnfgBA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-11T22:38:00.569030Z","bundle_sha256":"d136895c365355a2e7292b69823994738e0cb1f9449659d0fd0a1fdf6c353d7a"}}