{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:HNFYBUPPB3YD4YEJ7ZZM3M5YB5","short_pith_number":"pith:HNFYBUPP","canonical_record":{"source":{"id":"2605.14201","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-05-13T23:35:14Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"2a712b750a4df4ac839ae13ffa027d216726581c2912a98ff308894c2205dded","abstract_canon_sha256":"3320e439f5159f41c079f6c13015bf2c84d926fd2f7bf2a710bc6edbc20356e1"},"schema_version":"1.0"},"canonical_sha256":"3b4b80d1ef0ef03e6089fe72cdb3b80f7225ce5b5fd43dd3b69b3d13c3a556f8","source":{"kind":"arxiv","id":"2605.14201","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14201","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14201v1","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14201","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"pith_short_12","alias_value":"HNFYBUPPB3YD","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"HNFYBUPPB3YD4YEJ","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"HNFYBUPP","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:HNFYBUPPB3YD4YEJ7ZZM3M5YB5","target":"record","payload":{"canonical_record":{"source":{"id":"2605.14201","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-05-13T23:35:14Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"2a712b750a4df4ac839ae13ffa027d216726581c2912a98ff308894c2205dded","abstract_canon_sha256":"3320e439f5159f41c079f6c13015bf2c84d926fd2f7bf2a710bc6edbc20356e1"},"schema_version":"1.0"},"canonical_sha256":"3b4b80d1ef0ef03e6089fe72cdb3b80f7225ce5b5fd43dd3b69b3d13c3a556f8","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:11.040910Z","signature_b64":"W0GDnckzf2P2VGQq2PLZzzhU+pwFMDyAILHG8edYpWFDm1imYCFkYkoxL5LKeir4RUPj0Hrg+P8u114UYvcwBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3b4b80d1ef0ef03e6089fe72cdb3b80f7225ce5b5fd43dd3b69b3d13c3a556f8","last_reissued_at":"2026-05-17T23:39:11.040346Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:11.040346Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.14201","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ZNaeXOe3hPdTiqF14mP0q5m+xZv/K28e5RM7OuleyvrL1w8ZG4nCOiBtJzKi9LCvUJghI8nc2ymVE31zwF08Dw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-31T08:06:37.201084Z"},"content_sha256":"cb7d26d15dc7e63ce2af4b2b5213fb7c9e70925fa03716ae0d1346e8bff65aa0","schema_version":"1.0","event_id":"sha256:cb7d26d15dc7e63ce2af4b2b5213fb7c9e70925fa03716ae0d1346e8bff65aa0"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:HNFYBUPPB3YD4YEJ7ZZM3M5YB5","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"MAPLE: Latent Multi-Agent Play for End-to-End Autonomous Driving","license":"http://creativecommons.org/licenses/by/4.0/","headline":"MAPLE trains end-to-end driving models in closed loop using latent multi-agent rollouts without external simulators.","cross_cats":["cs.CV"],"primary_cat":"cs.RO","authors_text":"Deepti Hegde, Fatih Porikli, Hanno Ackermann, Hong Cai, Hsin-Pai Cheng, Litian Liu, Meysam Sadeghigooghari, Mohammad Ghavamzadeh, Pranav Desai, Rajeev Yasarla, Shizhong Han, Yunxiao Shi","submitted_at":"2026-05-13T23:35:14Z","abstract_excerpt":"Vision-language-action (VLA) models are effective as end-to-end motion planners, but can be brittle when evaluated in closed-loop settings due to being trained under traditional imitation learning framework. Existing closed-loop supervision approaches lack scalability and fail to completely model a reactive environment. We propose MAPLE, a novel framework for reactive, multi-agent rollout of a dynamic driving scenario in the latent space of the VLA model. The ego vehicle and nearby traffic agents are independently controlled over multi-step horizons, while being reactive to other agents in the"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"MAPLE achieves state-of-the-art driving performance on Bench2Drive and demonstrates scalable, closed-loop multi-agent play for robust E2E autonomous driving systems.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That latent-space rollouts of the VLA model accurately capture reactive multi-agent dynamics without external simulators or additional visual fidelity losses.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MAPLE performs closed-loop multi-agent training of VLA driving models entirely in latent space using supervised fine-tuning followed by RL with safety, progress, and diversity rewards, reaching SOTA on Bench2Drive.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"MAPLE trains end-to-end driving models in closed loop using latent multi-agent rollouts without external simulators.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e4ab8c227b13a2fb310cd3f72a62ef6ae136f34a07f8e2bdf70ca725091609e3"},"source":{"id":"2605.14201","kind":"arxiv","version":1},"verdict":{"id":"60512321-70b8-41b8-bcde-0cb0478f781f","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T04:38:17.797142Z","strongest_claim":"MAPLE achieves state-of-the-art driving performance on Bench2Drive and demonstrates scalable, closed-loop multi-agent play for robust E2E autonomous driving systems.","one_line_summary":"MAPLE performs closed-loop multi-agent training of VLA driving models entirely in latent space using supervised fine-tuning followed by RL with safety, progress, and diversity rewards, reaching SOTA on Bench2Drive.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That latent-space rollouts of the VLA model accurately capture reactive multi-agent dynamics without external simulators or additional visual fidelity losses.","pith_extraction_headline":"MAPLE trains end-to-end driving models in closed loop using latent multi-agent rollouts without external simulators."},"references":{"count":50,"sample":[{"doi":"","year":2020,"title":"Learning dexterous in-hand manipulation","work_id":"5d05e523-262c-439d-8e5c-8ae0702e4200","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond","work_id":"cbc2bb21-b6bb-46c0-80bf-107e195ffe10","ref_index":2,"cited_arxiv_id":"2308.12966","is_internal_anchor":true},{"doi":"","year":2025,"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","ref_index":3,"cited_arxiv_id":"2502.13923","is_internal_anchor":true},{"doi":"","year":2021,"title":"SimNet: Learning reactive self-driving simulations from real-world observations","work_id":"81616ec8-4a5d-4de7-adbe-2481bf60d85c","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Killian, Stuart Bowers, Ozan Sener, Philipp Kraehenbuehl, and Vladlen Koltun","work_id":"e3194ed7-9c32-41de-8af9-01be46d09252","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":50,"snapshot_sha256":"f4e783ae5822a921b0679a0cfed2ee2578f5a2a5e4f32509b117db376f9d2df3","internal_anchors":10},"formal_canon":{"evidence_count":2,"snapshot_sha256":"0ee1263a590e329dd168d7a7b86c960abf024d003553a249cb1013727ae618c1"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"60512321-70b8-41b8-bcde-0cb0478f781f"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"MTFQisdKOR7eTfKx/gUiUhfKL4pVx02r1VuW0mKkBE2A4/fgt11l5UnNpWKUNi90cwxjcssm//lBJ/Rb7xWUCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-31T08:06:37.201627Z"},"content_sha256":"aff49d05debdde95fe28ea57c461d98b770afe8456cbd663e6ecac029ab6e7e8","schema_version":"1.0","event_id":"sha256:aff49d05debdde95fe28ea57c461d98b770afe8456cbd663e6ecac029ab6e7e8"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/HNFYBUPPB3YD4YEJ7ZZM3M5YB5/bundle.json","state_url":"https://pith.science/pith/HNFYBUPPB3YD4YEJ7ZZM3M5YB5/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/HNFYBUPPB3YD4YEJ7ZZM3M5YB5/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":"2026-05-31T08:06:37Z","links":{"resolver":"https://pith.science/pith/HNFYBUPPB3YD4YEJ7ZZM3M5YB5","bundle":"https://pith.science/pith/HNFYBUPPB3YD4YEJ7ZZM3M5YB5/bundle.json","state":"https://pith.science/pith/HNFYBUPPB3YD4YEJ7ZZM3M5YB5/state.json","well_known_bundle":"https://pith.science/.well-known/pith/HNFYBUPPB3YD4YEJ7ZZM3M5YB5/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:HNFYBUPPB3YD4YEJ7ZZM3M5YB5","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":"3320e439f5159f41c079f6c13015bf2c84d926fd2f7bf2a710bc6edbc20356e1","cross_cats_sorted":["cs.CV"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-05-13T23:35:14Z","title_canon_sha256":"2a712b750a4df4ac839ae13ffa027d216726581c2912a98ff308894c2205dded"},"schema_version":"1.0","source":{"id":"2605.14201","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14201","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14201v1","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14201","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"pith_short_12","alias_value":"HNFYBUPPB3YD","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"HNFYBUPPB3YD4YEJ","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"HNFYBUPP","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:aff49d05debdde95fe28ea57c461d98b770afe8456cbd663e6ecac029ab6e7e8","target":"graph","created_at":"2026-05-17T23:39:11Z","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":"MAPLE achieves state-of-the-art driving performance on Bench2Drive and demonstrates scalable, closed-loop multi-agent play for robust E2E autonomous driving systems."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That latent-space rollouts of the VLA model accurately capture reactive multi-agent dynamics without external simulators or additional visual fidelity losses."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"MAPLE performs closed-loop multi-agent training of VLA driving models entirely in latent space using supervised fine-tuning followed by RL with safety, progress, and diversity rewards, reaching SOTA on Bench2Drive."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"MAPLE trains end-to-end driving models in closed loop using latent multi-agent rollouts without external simulators."}],"snapshot_sha256":"e4ab8c227b13a2fb310cd3f72a62ef6ae136f34a07f8e2bdf70ca725091609e3"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"0ee1263a590e329dd168d7a7b86c960abf024d003553a249cb1013727ae618c1"},"paper":{"abstract_excerpt":"Vision-language-action (VLA) models are effective as end-to-end motion planners, but can be brittle when evaluated in closed-loop settings due to being trained under traditional imitation learning framework. Existing closed-loop supervision approaches lack scalability and fail to completely model a reactive environment. We propose MAPLE, a novel framework for reactive, multi-agent rollout of a dynamic driving scenario in the latent space of the VLA model. The ego vehicle and nearby traffic agents are independently controlled over multi-step horizons, while being reactive to other agents in the","authors_text":"Deepti Hegde, Fatih Porikli, Hanno Ackermann, Hong Cai, Hsin-Pai Cheng, Litian Liu, Meysam Sadeghigooghari, Mohammad Ghavamzadeh, Pranav Desai, Rajeev Yasarla, Shizhong Han, Yunxiao Shi","cross_cats":["cs.CV"],"headline":"MAPLE trains end-to-end driving models in closed loop using latent multi-agent rollouts without external simulators.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-05-13T23:35:14Z","title":"MAPLE: Latent Multi-Agent Play for End-to-End Autonomous Driving"},"references":{"count":50,"internal_anchors":10,"resolved_work":50,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Learning dexterous in-hand manipulation","work_id":"5d05e523-262c-439d-8e5c-8ae0702e4200","year":2020},{"cited_arxiv_id":"2308.12966","doi":"","is_internal_anchor":true,"ref_index":2,"title":"Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond","work_id":"cbc2bb21-b6bb-46c0-80bf-107e195ffe10","year":2023},{"cited_arxiv_id":"2502.13923","doi":"","is_internal_anchor":true,"ref_index":3,"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"SimNet: Learning reactive self-driving simulations from real-world observations","work_id":"81616ec8-4a5d-4de7-adbe-2481bf60d85c","year":2021},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Killian, Stuart Bowers, Ozan Sener, Philipp Kraehenbuehl, and Vladlen Koltun","work_id":"e3194ed7-9c32-41de-8af9-01be46d09252","year":2025}],"snapshot_sha256":"f4e783ae5822a921b0679a0cfed2ee2578f5a2a5e4f32509b117db376f9d2df3"},"source":{"id":"2605.14201","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-15T04:38:17.797142Z","id":"60512321-70b8-41b8-bcde-0cb0478f781f","model_set":{"reader":"grok-4.3"},"one_line_summary":"MAPLE performs closed-loop multi-agent training of VLA driving models entirely in latent space using supervised fine-tuning followed by RL with safety, progress, and diversity rewards, reaching SOTA on Bench2Drive.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"MAPLE trains end-to-end driving models in closed loop using latent multi-agent rollouts without external simulators.","strongest_claim":"MAPLE achieves state-of-the-art driving performance on Bench2Drive and demonstrates scalable, closed-loop multi-agent play for robust E2E autonomous driving systems.","weakest_assumption":"That latent-space rollouts of the VLA model accurately capture reactive multi-agent dynamics without external simulators or additional visual fidelity losses."}},"verdict_id":"60512321-70b8-41b8-bcde-0cb0478f781f"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:cb7d26d15dc7e63ce2af4b2b5213fb7c9e70925fa03716ae0d1346e8bff65aa0","target":"record","created_at":"2026-05-17T23:39:11Z","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":"3320e439f5159f41c079f6c13015bf2c84d926fd2f7bf2a710bc6edbc20356e1","cross_cats_sorted":["cs.CV"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-05-13T23:35:14Z","title_canon_sha256":"2a712b750a4df4ac839ae13ffa027d216726581c2912a98ff308894c2205dded"},"schema_version":"1.0","source":{"id":"2605.14201","kind":"arxiv","version":1}},"canonical_sha256":"3b4b80d1ef0ef03e6089fe72cdb3b80f7225ce5b5fd43dd3b69b3d13c3a556f8","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3b4b80d1ef0ef03e6089fe72cdb3b80f7225ce5b5fd43dd3b69b3d13c3a556f8","first_computed_at":"2026-05-17T23:39:11.040346Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:11.040346Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"W0GDnckzf2P2VGQq2PLZzzhU+pwFMDyAILHG8edYpWFDm1imYCFkYkoxL5LKeir4RUPj0Hrg+P8u114UYvcwBA==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:11.040910Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.14201","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:cb7d26d15dc7e63ce2af4b2b5213fb7c9e70925fa03716ae0d1346e8bff65aa0","sha256:aff49d05debdde95fe28ea57c461d98b770afe8456cbd663e6ecac029ab6e7e8"],"state_sha256":"a13e8ff05ce4012b7f8c4ed3e633feb7e760480963424867a74da76cf088dad7"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"WQQ/bbrUVxNTv9js4hMKb792cH0wRIQPISoGVgnIUm2vuS8PDpFNfGWOFuzgPxvK/+b1XD29rIBKR5X0Jc9FBQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-31T08:06:37.205714Z","bundle_sha256":"4403c5494b17c9499306d7e804db708fdff1bf20deaf196fac18f26074463ed5"}}