{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:4PZ3Q5ZN2YGYZUXUX4FOVXUEXV","short_pith_number":"pith:4PZ3Q5ZN","schema_version":"1.0","canonical_sha256":"e3f3b8772dd60d8cd2f4bf0aeade84bd72331d1994a4cebff462c609d2bf4b6b","source":{"kind":"arxiv","id":"2605.24517","version":1},"attestation_state":"computed","paper":{"title":"ECHO: Terminal Agents Learn World Models for Free","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Ahmed Awadallah, Dimitris Papailiopoulos, Piero Kauffmann, Vaishnavi Shrivastava","submitted_at":"2026-05-23T11:08:46Z","abstract_excerpt":"CLI agents are the closest thing language models have to an embodied setting: the model emits commands, the terminal executes them, and the returned stream -- stdout, errors, files, logs, and traces -- records the consequences. We argue that this stream is a supervision signal, but standard agent RL discards it: GRPO-style training updates action tokens with sparse outcome-level rewards while ignoring environment responses already in the rollout. Failed rollouts provide little policy-gradient signal despite containing rich evidence about how the environment responds. We introduce ECHO (Environ"},"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":"2605.24517","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-23T11:08:46Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"1309c63a7251214fdda2f3f6840892c7209c0e1d1dc43acd0382922ae061c3b1","abstract_canon_sha256":"eda2dd69dfc4315b1cefe69d93316d874d3bf11df2f22c3030d2d44070d326bb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:03:44.013620Z","signature_b64":"1M08b8QYdIEgzWX2DXnHy1SrZpvxfKqDqsgX9yCto8LLMQqSoHh4uJJg54QVUnj0DYDk99MS6++Kh6AO3L2aCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e3f3b8772dd60d8cd2f4bf0aeade84bd72331d1994a4cebff462c609d2bf4b6b","last_reissued_at":"2026-05-26T01:03:44.012860Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:03:44.012860Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ECHO: Terminal Agents Learn World Models for Free","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Ahmed Awadallah, Dimitris Papailiopoulos, Piero Kauffmann, Vaishnavi Shrivastava","submitted_at":"2026-05-23T11:08:46Z","abstract_excerpt":"CLI agents are the closest thing language models have to an embodied setting: the model emits commands, the terminal executes them, and the returned stream -- stdout, errors, files, logs, and traces -- records the consequences. We argue that this stream is a supervision signal, but standard agent RL discards it: GRPO-style training updates action tokens with sparse outcome-level rewards while ignoring environment responses already in the rollout. Failed rollouts provide little policy-gradient signal despite containing rich evidence about how the environment responds. We introduce ECHO (Environ"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.24517","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/2605.24517/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":"2605.24517","created_at":"2026-05-26T01:03:44.012983+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.24517v1","created_at":"2026-05-26T01:03:44.012983+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.24517","created_at":"2026-05-26T01:03:44.012983+00:00"},{"alias_kind":"pith_short_12","alias_value":"4PZ3Q5ZN2YGY","created_at":"2026-05-26T01:03:44.012983+00:00"},{"alias_kind":"pith_short_16","alias_value":"4PZ3Q5ZN2YGYZUXU","created_at":"2026-05-26T01:03:44.012983+00:00"},{"alias_kind":"pith_short_8","alias_value":"4PZ3Q5ZN","created_at":"2026-05-26T01:03:44.012983+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/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV","json":"https://pith.science/pith/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV.json","graph_json":"https://pith.science/api/pith-number/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV/graph.json","events_json":"https://pith.science/api/pith-number/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV/events.json","paper":"https://pith.science/paper/4PZ3Q5ZN"},"agent_actions":{"view_html":"https://pith.science/pith/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV","download_json":"https://pith.science/pith/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV.json","view_paper":"https://pith.science/paper/4PZ3Q5ZN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.24517&json=true","fetch_graph":"https://pith.science/api/pith-number/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV/graph.json","fetch_events":"https://pith.science/api/pith-number/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV/action/storage_attestation","attest_author":"https://pith.science/pith/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV/action/author_attestation","sign_citation":"https://pith.science/pith/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV/action/citation_signature","submit_replication":"https://pith.science/pith/4PZ3Q5ZN2YGYZUXUX4FOVXUEXV/action/replication_record"}},"created_at":"2026-05-26T01:03:44.012983+00:00","updated_at":"2026-05-26T01:03:44.012983+00:00"}