{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:GZ6GYSFJZCGCZRRVAOLOEIACB2","short_pith_number":"pith:GZ6GYSFJ","schema_version":"1.0","canonical_sha256":"367c6c48a9c88c2cc6350396e220020eacce41680c36723a250b048554e551de","source":{"kind":"arxiv","id":"2606.26175","version":1},"attestation_state":"computed","paper":{"title":"RMTL: Reinforced Micro-task Learning for Long-Horizon Manipulation with VLM Rewards","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.RO","authors_text":"An{\\i}l Can Ate\\c{s}, Cihan Topal, Orhan Kahraman","submitted_at":"2026-06-24T10:32:48Z","abstract_excerpt":"Reinforcement learning (RL) for robotic manipulation often requires manually designing a dense reward function, which is difficult to tune and often fragile, or learning a reward from human demonstrations or preferences, which can be expensive. A recent line of work uses pretrained vision-language models (VLMs) as zero-shot reward models, replacing these costs with a single text prompt. However, we argue that a single global prompt is too coarse for long-horizon manipulation tasks with randomized initial conditions. The single-prompt VLM reward is near-flat for much of the trajectory, making e"},"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":false},"canonical_record":{"source":{"id":"2606.26175","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2026-06-24T10:32:48Z","cross_cats_sorted":[],"title_canon_sha256":"a3d256e0d7716201718f02ed07ed7b58053d3f1e15d872d05cc904e2c9762c66","abstract_canon_sha256":"22e241f9b63e474c6f4676415485681474f32d0993edbf010acd86f037648d81"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-26T00:15:28.524527Z","signature_b64":"2pmGHuifnRODRJRPJ9/rV7UFuY5Z6dG1tw0iHt84nM5tX1z6mdgTQPunekhQwezEgDeXLTS+o01vKoBXzhfdBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"367c6c48a9c88c2cc6350396e220020eacce41680c36723a250b048554e551de","last_reissued_at":"2026-06-26T00:15:28.524116Z","signature_status":"signed_v1","first_computed_at":"2026-06-26T00:15:28.524116Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RMTL: Reinforced Micro-task Learning for Long-Horizon Manipulation with VLM Rewards","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.RO","authors_text":"An{\\i}l Can Ate\\c{s}, Cihan Topal, Orhan Kahraman","submitted_at":"2026-06-24T10:32:48Z","abstract_excerpt":"Reinforcement learning (RL) for robotic manipulation often requires manually designing a dense reward function, which is difficult to tune and often fragile, or learning a reward from human demonstrations or preferences, which can be expensive. A recent line of work uses pretrained vision-language models (VLMs) as zero-shot reward models, replacing these costs with a single text prompt. However, we argue that a single global prompt is too coarse for long-horizon manipulation tasks with randomized initial conditions. The single-prompt VLM reward is near-flat for much of the trajectory, making e"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.26175","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2606.26175/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":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"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":"2606.26175","created_at":"2026-06-26T00:15:28.524172+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.26175v1","created_at":"2026-06-26T00:15:28.524172+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.26175","created_at":"2026-06-26T00:15:28.524172+00:00"},{"alias_kind":"pith_short_12","alias_value":"GZ6GYSFJZCGC","created_at":"2026-06-26T00:15:28.524172+00:00"},{"alias_kind":"pith_short_16","alias_value":"GZ6GYSFJZCGCZRRV","created_at":"2026-06-26T00:15:28.524172+00:00"},{"alias_kind":"pith_short_8","alias_value":"GZ6GYSFJ","created_at":"2026-06-26T00:15:28.524172+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GZ6GYSFJZCGCZRRVAOLOEIACB2","json":"https://pith.science/pith/GZ6GYSFJZCGCZRRVAOLOEIACB2.json","graph_json":"https://pith.science/api/pith-number/GZ6GYSFJZCGCZRRVAOLOEIACB2/graph.json","events_json":"https://pith.science/api/pith-number/GZ6GYSFJZCGCZRRVAOLOEIACB2/events.json","paper":"https://pith.science/paper/GZ6GYSFJ"},"agent_actions":{"view_html":"https://pith.science/pith/GZ6GYSFJZCGCZRRVAOLOEIACB2","download_json":"https://pith.science/pith/GZ6GYSFJZCGCZRRVAOLOEIACB2.json","view_paper":"https://pith.science/paper/GZ6GYSFJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.26175&json=true","fetch_graph":"https://pith.science/api/pith-number/GZ6GYSFJZCGCZRRVAOLOEIACB2/graph.json","fetch_events":"https://pith.science/api/pith-number/GZ6GYSFJZCGCZRRVAOLOEIACB2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GZ6GYSFJZCGCZRRVAOLOEIACB2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GZ6GYSFJZCGCZRRVAOLOEIACB2/action/storage_attestation","attest_author":"https://pith.science/pith/GZ6GYSFJZCGCZRRVAOLOEIACB2/action/author_attestation","sign_citation":"https://pith.science/pith/GZ6GYSFJZCGCZRRVAOLOEIACB2/action/citation_signature","submit_replication":"https://pith.science/pith/GZ6GYSFJZCGCZRRVAOLOEIACB2/action/replication_record"}},"created_at":"2026-06-26T00:15:28.524172+00:00","updated_at":"2026-06-26T00:15:28.524172+00:00"}