{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:JO2OSSKHNSUX3EPGK72F2A7S2B","short_pith_number":"pith:JO2OSSKH","canonical_record":{"source":{"id":"2603.25340","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-03-26T11:30:44Z","cross_cats_sorted":[],"title_canon_sha256":"b81dec45054f97d889dd3e9fc8e6a6c4b10f0b70d449fb36e98d42bc40260256","abstract_canon_sha256":"0cf7210c30b7c99e5301fe6099b8f4977c9a456a51c925172c937b42dfe89f26"},"schema_version":"1.0"},"canonical_sha256":"4bb4e949476ca97d91e657f45d03f2d06a74156ef7ae405b0b812d0b1b2d5b9c","source":{"kind":"arxiv","id":"2603.25340","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2603.25340","created_at":"2026-05-18T03:09:22Z"},{"alias_kind":"arxiv_version","alias_value":"2603.25340v2","created_at":"2026-05-18T03:09:22Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.25340","created_at":"2026-05-18T03:09:22Z"},{"alias_kind":"pith_short_12","alias_value":"JO2OSSKHNSUX","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"JO2OSSKHNSUX3EPG","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"JO2OSSKH","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:JO2OSSKHNSUX3EPGK72F2A7S2B","target":"record","payload":{"canonical_record":{"source":{"id":"2603.25340","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-03-26T11:30:44Z","cross_cats_sorted":[],"title_canon_sha256":"b81dec45054f97d889dd3e9fc8e6a6c4b10f0b70d449fb36e98d42bc40260256","abstract_canon_sha256":"0cf7210c30b7c99e5301fe6099b8f4977c9a456a51c925172c937b42dfe89f26"},"schema_version":"1.0"},"canonical_sha256":"4bb4e949476ca97d91e657f45d03f2d06a74156ef7ae405b0b812d0b1b2d5b9c","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T03:09:22.532188Z","signature_b64":"HIPQ+/dXBQyg/TgzGO1i90+MXyJUStjJYSED3IXEwcSZaTODAt5/iopkl5SC9q4HjN7CXq7Pe0bdikm7ITTHAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4bb4e949476ca97d91e657f45d03f2d06a74156ef7ae405b0b812d0b1b2d5b9c","last_reissued_at":"2026-05-18T03:09:22.531461Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T03:09:22.531461Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2603.25340","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T03:09:22Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"0XFX0qpELUjSjNHFZnoSPideFFSrMQA3tMP/st/wcobAOhl9Bfnrz2ZJ/faW46e99lq6grqhproq54XC0eW7DQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-19T20:46:14.306139Z"},"content_sha256":"f7c3c8ea96a9b117edb459c71475aee07df2081257e286eab38bdf15d2fa2b74","schema_version":"1.0","event_id":"sha256:f7c3c8ea96a9b117edb459c71475aee07df2081257e286eab38bdf15d2fa2b74"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:JO2OSSKHNSUX3EPGK72F2A7S2B","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Large Language Model as Token Compressor and Decompressor","license":"http://creativecommons.org/licenses/by/4.0/","headline":"An off-the-shelf LLM can be fine-tuned with LoRA to compress long texts into adaptive sequences of Z-tokens while preserving reconstruction and task performance.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jielei Zhang, Junkai Lin, Tianhao Zhao, Wei Yang, Wenbing Li, Yiran Wang, Zikai Song","submitted_at":"2026-03-26T11:30:44Z","abstract_excerpt":"In this paper, we study whether an off-the-shelf LLM can be adapted into a discrete, variable-length token compressor and decompressor for long-context processing. To this end, we design a self-expressive autoencoding framework that fine-tunes a pretrained LLM with lightweight LoRA adapters to map long texts into compact sequences of learned latent codes, termed Z-tokens, and to decode them back into natural language or task outputs. The resulting representation is content-adaptive: less predictable or information-dense segments can receive more Z-tokens, while redundant regions can be represe"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"an off-the-shelf LLM can be adapted into a discrete, variable-length token compressor and decompressor for long-context processing","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"that fine-tuning with LoRA on the self-expressive autoencoding objective will produce Z-tokens that preserve enough information for both faithful reconstruction and downstream task performance without requiring extensive post-hoc adjustments","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A pretrained LLM is adapted via LoRA fine-tuning into a content-adaptive compressor that maps long texts to compact variable-length Z-token sequences while preserving reconstruction quality and downstream performance.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"An off-the-shelf LLM can be fine-tuned with LoRA to compress long texts into adaptive sequences of Z-tokens while preserving reconstruction and task performance.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ab74a9d5a7564ab3a0b542b623027b4812b61d5a48311c23e4afd59113ab3f18"},"source":{"id":"2603.25340","kind":"arxiv","version":2},"verdict":{"id":"aeeb3b15-b7ea-4d5e-b3d9-aa7f68a69caf","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T00:51:31.603008Z","strongest_claim":"an off-the-shelf LLM can be adapted into a discrete, variable-length token compressor and decompressor for long-context processing","one_line_summary":"A pretrained LLM is adapted via LoRA fine-tuning into a content-adaptive compressor that maps long texts to compact variable-length Z-token sequences while preserving reconstruction quality and downstream performance.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"that fine-tuning with LoRA on the self-expressive autoencoding objective will produce Z-tokens that preserve enough information for both faithful reconstruction and downstream task performance without requiring extensive post-hoc adjustments","pith_extraction_headline":"An off-the-shelf LLM can be fine-tuned with LoRA to compress long texts into adaptive sequences of Z-tokens while preserving reconstruction and task performance."},"references":{"count":47,"sample":[{"doi":"","year":2020,"title":"Peters, and Arman Cohan","work_id":"477d1550-5247-4659-85a2-1cc846c63115","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Token merging: Your vit but faster, 2023","work_id":"790954c9-35d8-4397-92ae-24d94c563cc4","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Hudson, Ehsan Adeli, et al","work_id":"10e2738f-b505-4455-824c-58c06c17d125","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Sub- biah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakan- tan, Pranav Shyam, Girish Sastry, Amanda Askell, Sand- hini Agarwal, Ariel Herbert-V oss, ","work_id":"817f656c-a9d7-4ab2-a333-8160212f136f","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Adapting language models to compress con- texts, 2023","work_id":"b022c61c-4cd4-40d2-b589-dbb244102410","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":47,"snapshot_sha256":"2692c8234c92c8b3dd6d8180afe3b3c8ef62bd5294d4beb60c953b75c416b154","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":"aeeb3b15-b7ea-4d5e-b3d9-aa7f68a69caf"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T03:09:22Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"kHXMAkfBvFb8/osIVHLEQSb2m/YncfqUnA2yC7vplR5nQRa+CrhB3XE4OOsfaHrThNIzZJ7P5waotF1bxioSCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-19T20:46:14.306688Z"},"content_sha256":"c285306664dd26e14ff4479d98d4a62bb9dbdb9246f2a7055be5ac8834b06546","schema_version":"1.0","event_id":"sha256:c285306664dd26e14ff4479d98d4a62bb9dbdb9246f2a7055be5ac8834b06546"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/JO2OSSKHNSUX3EPGK72F2A7S2B/bundle.json","state_url":"https://pith.science/pith/JO2OSSKHNSUX3EPGK72F2A7S2B/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/JO2OSSKHNSUX3EPGK72F2A7S2B/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":"2026-05-19T20:46:14Z","links":{"resolver":"https://pith.science/pith/JO2OSSKHNSUX3EPGK72F2A7S2B","bundle":"https://pith.science/pith/JO2OSSKHNSUX3EPGK72F2A7S2B/bundle.json","state":"https://pith.science/pith/JO2OSSKHNSUX3EPGK72F2A7S2B/state.json","well_known_bundle":"https://pith.science/.well-known/pith/JO2OSSKHNSUX3EPGK72F2A7S2B/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:JO2OSSKHNSUX3EPGK72F2A7S2B","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":"0cf7210c30b7c99e5301fe6099b8f4977c9a456a51c925172c937b42dfe89f26","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-03-26T11:30:44Z","title_canon_sha256":"b81dec45054f97d889dd3e9fc8e6a6c4b10f0b70d449fb36e98d42bc40260256"},"schema_version":"1.0","source":{"id":"2603.25340","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2603.25340","created_at":"2026-05-18T03:09:22Z"},{"alias_kind":"arxiv_version","alias_value":"2603.25340v2","created_at":"2026-05-18T03:09:22Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.25340","created_at":"2026-05-18T03:09:22Z"},{"alias_kind":"pith_short_12","alias_value":"JO2OSSKHNSUX","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"JO2OSSKHNSUX3EPG","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"JO2OSSKH","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:c285306664dd26e14ff4479d98d4a62bb9dbdb9246f2a7055be5ac8834b06546","target":"graph","created_at":"2026-05-18T03:09:22Z","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":"an off-the-shelf LLM can be adapted into a discrete, variable-length token compressor and decompressor for long-context processing"},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"that fine-tuning with LoRA on the self-expressive autoencoding objective will produce Z-tokens that preserve enough information for both faithful reconstruction and downstream task performance without requiring extensive post-hoc adjustments"},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"A pretrained LLM is adapted via LoRA fine-tuning into a content-adaptive compressor that maps long texts to compact variable-length Z-token sequences while preserving reconstruction quality and downstream performance."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"An off-the-shelf LLM can be fine-tuned with LoRA to compress long texts into adaptive sequences of Z-tokens while preserving reconstruction and task performance."}],"snapshot_sha256":"ab74a9d5a7564ab3a0b542b623027b4812b61d5a48311c23e4afd59113ab3f18"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"In this paper, we study whether an off-the-shelf LLM can be adapted into a discrete, variable-length token compressor and decompressor for long-context processing. To this end, we design a self-expressive autoencoding framework that fine-tunes a pretrained LLM with lightweight LoRA adapters to map long texts into compact sequences of learned latent codes, termed Z-tokens, and to decode them back into natural language or task outputs. The resulting representation is content-adaptive: less predictable or information-dense segments can receive more Z-tokens, while redundant regions can be represe","authors_text":"Jielei Zhang, Junkai Lin, Tianhao Zhao, Wei Yang, Wenbing Li, Yiran Wang, Zikai Song","cross_cats":[],"headline":"An off-the-shelf LLM can be fine-tuned with LoRA to compress long texts into adaptive sequences of Z-tokens while preserving reconstruction and task performance.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-03-26T11:30:44Z","title":"Large Language Model as Token Compressor and Decompressor"},"references":{"count":47,"internal_anchors":0,"resolved_work":47,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Peters, and Arman Cohan","work_id":"477d1550-5247-4659-85a2-1cc846c63115","year":2020},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Token merging: Your vit but faster, 2023","work_id":"790954c9-35d8-4397-92ae-24d94c563cc4","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Hudson, Ehsan Adeli, et al","work_id":"10e2738f-b505-4455-824c-58c06c17d125","year":2022},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Sub- biah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakan- tan, Pranav Shyam, Girish Sastry, Amanda Askell, Sand- hini Agarwal, Ariel Herbert-V oss, ","work_id":"817f656c-a9d7-4ab2-a333-8160212f136f","year":2020},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Adapting language models to compress con- texts, 2023","work_id":"b022c61c-4cd4-40d2-b589-dbb244102410","year":2023}],"snapshot_sha256":"2692c8234c92c8b3dd6d8180afe3b3c8ef62bd5294d4beb60c953b75c416b154"},"source":{"id":"2603.25340","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-15T00:51:31.603008Z","id":"aeeb3b15-b7ea-4d5e-b3d9-aa7f68a69caf","model_set":{"reader":"grok-4.3"},"one_line_summary":"A pretrained LLM is adapted via LoRA fine-tuning into a content-adaptive compressor that maps long texts to compact variable-length Z-token sequences while preserving reconstruction quality and downstream performance.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"An off-the-shelf LLM can be fine-tuned with LoRA to compress long texts into adaptive sequences of Z-tokens while preserving reconstruction and task performance.","strongest_claim":"an off-the-shelf LLM can be adapted into a discrete, variable-length token compressor and decompressor for long-context processing","weakest_assumption":"that fine-tuning with LoRA on the self-expressive autoencoding objective will produce Z-tokens that preserve enough information for both faithful reconstruction and downstream task performance without requiring extensive post-hoc adjustments"}},"verdict_id":"aeeb3b15-b7ea-4d5e-b3d9-aa7f68a69caf"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:f7c3c8ea96a9b117edb459c71475aee07df2081257e286eab38bdf15d2fa2b74","target":"record","created_at":"2026-05-18T03:09:22Z","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":"0cf7210c30b7c99e5301fe6099b8f4977c9a456a51c925172c937b42dfe89f26","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-03-26T11:30:44Z","title_canon_sha256":"b81dec45054f97d889dd3e9fc8e6a6c4b10f0b70d449fb36e98d42bc40260256"},"schema_version":"1.0","source":{"id":"2603.25340","kind":"arxiv","version":2}},"canonical_sha256":"4bb4e949476ca97d91e657f45d03f2d06a74156ef7ae405b0b812d0b1b2d5b9c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4bb4e949476ca97d91e657f45d03f2d06a74156ef7ae405b0b812d0b1b2d5b9c","first_computed_at":"2026-05-18T03:09:22.531461Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T03:09:22.531461Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"HIPQ+/dXBQyg/TgzGO1i90+MXyJUStjJYSED3IXEwcSZaTODAt5/iopkl5SC9q4HjN7CXq7Pe0bdikm7ITTHAQ==","signature_status":"signed_v1","signed_at":"2026-05-18T03:09:22.532188Z","signed_message":"canonical_sha256_bytes"},"source_id":"2603.25340","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f7c3c8ea96a9b117edb459c71475aee07df2081257e286eab38bdf15d2fa2b74","sha256:c285306664dd26e14ff4479d98d4a62bb9dbdb9246f2a7055be5ac8834b06546"],"state_sha256":"30552067ad2f5bd0ed3d4c947c0a2751eb42b13f1deb6621652e0f6a785a502b"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ae4ThR6mXuPwqjcbxI9PUCLlFlZsGk459HmCARrRYkW2k8bHn7CiXLAmwFBvrb/tIKL9CEs/XCL24xql+/uPCQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-19T20:46:14.309300Z","bundle_sha256":"2706a57271b1b478bb0b99355610d39a63df917916e5ecde10271c084dade3c1"}}