{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:ADUZIQ2IT4Y53I7FZWGL7RDVEE","short_pith_number":"pith:ADUZIQ2I","canonical_record":{"source":{"id":"2605.11212","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-11T20:27:54Z","cross_cats_sorted":[],"title_canon_sha256":"d800b8ce1fee17a89f55d674060e49f0380e0fbdfd8a61325bb92fd03d953f19","abstract_canon_sha256":"f02f78b173f689067b109e15afd28d0da8778a3d70b888e35f85d6da381654c8"},"schema_version":"1.0"},"canonical_sha256":"00e99443489f31dda3e5cd8cbfc475212e54e7cf43d283b661d28d31690576ea","source":{"kind":"arxiv","id":"2605.11212","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.11212","created_at":"2026-06-09T01:04:43Z"},{"alias_kind":"arxiv_version","alias_value":"2605.11212v3","created_at":"2026-06-09T01:04:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.11212","created_at":"2026-06-09T01:04:43Z"},{"alias_kind":"pith_short_12","alias_value":"ADUZIQ2IT4Y5","created_at":"2026-06-09T01:04:43Z"},{"alias_kind":"pith_short_16","alias_value":"ADUZIQ2IT4Y53I7F","created_at":"2026-06-09T01:04:43Z"},{"alias_kind":"pith_short_8","alias_value":"ADUZIQ2I","created_at":"2026-06-09T01:04:43Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:ADUZIQ2IT4Y53I7FZWGL7RDVEE","target":"record","payload":{"canonical_record":{"source":{"id":"2605.11212","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-11T20:27:54Z","cross_cats_sorted":[],"title_canon_sha256":"d800b8ce1fee17a89f55d674060e49f0380e0fbdfd8a61325bb92fd03d953f19","abstract_canon_sha256":"f02f78b173f689067b109e15afd28d0da8778a3d70b888e35f85d6da381654c8"},"schema_version":"1.0"},"canonical_sha256":"00e99443489f31dda3e5cd8cbfc475212e54e7cf43d283b661d28d31690576ea","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T01:04:43.528011Z","signature_b64":"WHIFJBFNYQ0Q/kEIAAagdvoHJ8/YNUPZMat3UU2TxXVCLh2kaHsmXAsuj1GEfCn0z+dGDCufNR+warsp39xMAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"00e99443489f31dda3e5cd8cbfc475212e54e7cf43d283b661d28d31690576ea","last_reissued_at":"2026-06-09T01:04:43.527598Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T01:04:43.527598Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.11212","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":"2026-06-09T01:04:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"iqTCgIhQzzzEO4YQ7vPFQdnE/5hKg9q51uz+IcF+fQGM+v6bAbAmAT57VA4kl2fqc4L10kzCGbTJexUWEUBfCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-29T13:14:16.846975Z"},"content_sha256":"2214216b8a928f24ae4e1932c94eafc0e98c4d123c37d3562482eb28bdba9b3e","schema_version":"1.0","event_id":"sha256:2214216b8a928f24ae4e1932c94eafc0e98c4d123c37d3562482eb28bdba9b3e"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:ADUZIQ2IT4Y53I7FZWGL7RDVEE","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"ReVision: Scaling Computer-Use Agents via Temporal Visual Redundancy Reduction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"ReVision removes redundant visual patches from agent history screenshots to cut token usage by 46 percent while raising success rates by 3 percent.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amirhossein Abaskohi, Giuseppe Carenini, Peter West, Pranit Chawla, Vibhav Vineet, Yuhang He","submitted_at":"2026-05-11T20:27:54Z","abstract_excerpt":"Computer-use agents (CUAs) rely on visual observations of graphical user interfaces, where each screenshot is encoded into a large number of visual tokens. As interaction trajectories grow, the token cost increases rapidly, limiting the amount of history that can be incorporated under fixed context and compute budgets. This has resulted in no or very limited improvement in the performance when using history unlike other domains. We address this inefficiency by introducing ReVision, which is used to train multimodal language models on trajectories where redundant visual patches are removed usin"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"when processing trajectories with 5 history screenshots using Qwen2.5-VL-7B, ReVision reduces token usage by approximately 46% on average while improving success rate by 3% over the no drop baseline. This establishes a clear efficiency gain, enabling agents to process longer trajectories with fewer tokens.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The learned patch selector accurately identifies and removes only redundant patches without discarding task-critical visual information required for correct agent actions.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ReVision reduces visual token usage by 46% on average in agent trajectories via a learned patch selector and improves success rates by 3% on three benchmarks, showing that history saturation stems from inefficient representations rather than lack of utility.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ReVision removes redundant visual patches from agent history screenshots to cut token usage by 46 percent while raising success rates by 3 percent.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"37f27a6f6453c8c88c5b8806d0834c33de99cf167c055eccfb2f761b83caa9b4"},"source":{"id":"2605.11212","kind":"arxiv","version":3},"verdict":{"id":"1339481c-c851-4c5a-a752-b812463d0ba1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T20:52:25.854200Z","strongest_claim":"when processing trajectories with 5 history screenshots using Qwen2.5-VL-7B, ReVision reduces token usage by approximately 46% on average while improving success rate by 3% over the no drop baseline. This establishes a clear efficiency gain, enabling agents to process longer trajectories with fewer tokens.","one_line_summary":"ReVision reduces visual token usage by 46% on average in agent trajectories via a learned patch selector and improves success rates by 3% on three benchmarks, showing that history saturation stems from inefficient representations rather than lack of utility.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The learned patch selector accurately identifies and removes only redundant patches without discarding task-critical visual information required for correct agent actions.","pith_extraction_headline":"ReVision removes redundant visual patches from agent history screenshots to cut token usage by 46 percent while raising success rates by 3 percent."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.11212/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"claim_evidence","ran_at":"2026-05-20T04:42:00.910216Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T12:40:14.363956Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T10:01:17.491965Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T08:40:06.839834Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"d3c7c083817a71eecc3049767b633d0ba7163c174b43b493db3a54b6f5f025d9"},"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":"1339481c-c851-4c5a-a752-b812463d0ba1"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-09T01:04:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"lylj4wPD5I9DSTllsaSgtMCY+ePYSwULN7XvyraTD4iUCZVlTq4gNIIr5+piD6TK325m29OxTsR4lKsEEr1RCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-29T13:14:16.847482Z"},"content_sha256":"f884bd5fa0e4910d7f9e6039ffca771429084b8ea88f45fa78f08766dbe68175","schema_version":"1.0","event_id":"sha256:f884bd5fa0e4910d7f9e6039ffca771429084b8ea88f45fa78f08766dbe68175"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/ADUZIQ2IT4Y53I7FZWGL7RDVEE/bundle.json","state_url":"https://pith.science/pith/ADUZIQ2IT4Y53I7FZWGL7RDVEE/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/ADUZIQ2IT4Y53I7FZWGL7RDVEE/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":"2026-06-29T13:14:16Z","links":{"resolver":"https://pith.science/pith/ADUZIQ2IT4Y53I7FZWGL7RDVEE","bundle":"https://pith.science/pith/ADUZIQ2IT4Y53I7FZWGL7RDVEE/bundle.json","state":"https://pith.science/pith/ADUZIQ2IT4Y53I7FZWGL7RDVEE/state.json","well_known_bundle":"https://pith.science/.well-known/pith/ADUZIQ2IT4Y53I7FZWGL7RDVEE/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:ADUZIQ2IT4Y53I7FZWGL7RDVEE","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":"f02f78b173f689067b109e15afd28d0da8778a3d70b888e35f85d6da381654c8","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-11T20:27:54Z","title_canon_sha256":"d800b8ce1fee17a89f55d674060e49f0380e0fbdfd8a61325bb92fd03d953f19"},"schema_version":"1.0","source":{"id":"2605.11212","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.11212","created_at":"2026-06-09T01:04:43Z"},{"alias_kind":"arxiv_version","alias_value":"2605.11212v3","created_at":"2026-06-09T01:04:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.11212","created_at":"2026-06-09T01:04:43Z"},{"alias_kind":"pith_short_12","alias_value":"ADUZIQ2IT4Y5","created_at":"2026-06-09T01:04:43Z"},{"alias_kind":"pith_short_16","alias_value":"ADUZIQ2IT4Y53I7F","created_at":"2026-06-09T01:04:43Z"},{"alias_kind":"pith_short_8","alias_value":"ADUZIQ2I","created_at":"2026-06-09T01:04:43Z"}],"graph_snapshots":[{"event_id":"sha256:f884bd5fa0e4910d7f9e6039ffca771429084b8ea88f45fa78f08766dbe68175","target":"graph","created_at":"2026-06-09T01:04:43Z","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":"when processing trajectories with 5 history screenshots using Qwen2.5-VL-7B, ReVision reduces token usage by approximately 46% on average while improving success rate by 3% over the no drop baseline. This establishes a clear efficiency gain, enabling agents to process longer trajectories with fewer tokens."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The learned patch selector accurately identifies and removes only redundant patches without discarding task-critical visual information required for correct agent actions."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"ReVision reduces visual token usage by 46% on average in agent trajectories via a learned patch selector and improves success rates by 3% on three benchmarks, showing that history saturation stems from inefficient representations rather than lack of utility."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"ReVision removes redundant visual patches from agent history screenshots to cut token usage by 46 percent while raising success rates by 3 percent."}],"snapshot_sha256":"37f27a6f6453c8c88c5b8806d0834c33de99cf167c055eccfb2f761b83caa9b4"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-20T04:42:00.910216Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T12:40:14.363956Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T10:01:17.491965Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T08:40:06.839834Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.11212/integrity.json","findings":[],"snapshot_sha256":"d3c7c083817a71eecc3049767b633d0ba7163c174b43b493db3a54b6f5f025d9","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Computer-use agents (CUAs) rely on visual observations of graphical user interfaces, where each screenshot is encoded into a large number of visual tokens. As interaction trajectories grow, the token cost increases rapidly, limiting the amount of history that can be incorporated under fixed context and compute budgets. This has resulted in no or very limited improvement in the performance when using history unlike other domains. We address this inefficiency by introducing ReVision, which is used to train multimodal language models on trajectories where redundant visual patches are removed usin","authors_text":"Amirhossein Abaskohi, Giuseppe Carenini, Peter West, Pranit Chawla, Vibhav Vineet, Yuhang He","cross_cats":[],"headline":"ReVision removes redundant visual patches from agent history screenshots to cut token usage by 46 percent while raising success rates by 3 percent.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-11T20:27:54Z","title":"ReVision: Scaling Computer-Use Agents via Temporal Visual Redundancy Reduction"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.11212","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-14T20:52:25.854200Z","id":"1339481c-c851-4c5a-a752-b812463d0ba1","model_set":{"reader":"grok-4.3"},"one_line_summary":"ReVision reduces visual token usage by 46% on average in agent trajectories via a learned patch selector and improves success rates by 3% on three benchmarks, showing that history saturation stems from inefficient representations rather than lack of utility.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"ReVision removes redundant visual patches from agent history screenshots to cut token usage by 46 percent while raising success rates by 3 percent.","strongest_claim":"when processing trajectories with 5 history screenshots using Qwen2.5-VL-7B, ReVision reduces token usage by approximately 46% on average while improving success rate by 3% over the no drop baseline. This establishes a clear efficiency gain, enabling agents to process longer trajectories with fewer tokens.","weakest_assumption":"The learned patch selector accurately identifies and removes only redundant patches without discarding task-critical visual information required for correct agent actions."}},"verdict_id":"1339481c-c851-4c5a-a752-b812463d0ba1"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:2214216b8a928f24ae4e1932c94eafc0e98c4d123c37d3562482eb28bdba9b3e","target":"record","created_at":"2026-06-09T01:04:43Z","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":"f02f78b173f689067b109e15afd28d0da8778a3d70b888e35f85d6da381654c8","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-11T20:27:54Z","title_canon_sha256":"d800b8ce1fee17a89f55d674060e49f0380e0fbdfd8a61325bb92fd03d953f19"},"schema_version":"1.0","source":{"id":"2605.11212","kind":"arxiv","version":3}},"canonical_sha256":"00e99443489f31dda3e5cd8cbfc475212e54e7cf43d283b661d28d31690576ea","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"00e99443489f31dda3e5cd8cbfc475212e54e7cf43d283b661d28d31690576ea","first_computed_at":"2026-06-09T01:04:43.527598Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-09T01:04:43.527598Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"WHIFJBFNYQ0Q/kEIAAagdvoHJ8/YNUPZMat3UU2TxXVCLh2kaHsmXAsuj1GEfCn0z+dGDCufNR+warsp39xMAw==","signature_status":"signed_v1","signed_at":"2026-06-09T01:04:43.528011Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.11212","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:2214216b8a928f24ae4e1932c94eafc0e98c4d123c37d3562482eb28bdba9b3e","sha256:f884bd5fa0e4910d7f9e6039ffca771429084b8ea88f45fa78f08766dbe68175"],"state_sha256":"b071bffe4ad28f3ab00fc9d441b93f1a0704f94e2ef14aeae05626070a58acbc"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ST7araWU0qsymlHDDumHOrkdHlUHrPT5MSn4tqbDEVTwsOg6SBSuGS7iIJXtI24Tz5ifzKsiSpR7jSz65T7vDQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-29T13:14:16.849717Z","bundle_sha256":"1d8897f18457552cdb9abebc4a4237777ac44e10f52dd6e26869fa59eec39605"}}