{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:B7XDWRI47A3YI3JUJJVZDT5III","short_pith_number":"pith:B7XDWRI4","canonical_record":{"source":{"id":"2605.08738","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T06:50:35Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"d690173b393dfc6bc5926ee52dd327d38501a6e48b0392e1e7e4e6f3152c0c84","abstract_canon_sha256":"dd75df23f929b8bad3a1878c02869cdebdf0033ee3e365b579a310e61361a38d"},"schema_version":"1.0"},"canonical_sha256":"0fee3b451cf837846d344a6b91cfa8423e2a12641676538d0577db1ea14f41df","source":{"kind":"arxiv","id":"2605.08738","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.08738","created_at":"2026-05-20T00:04:34Z"},{"alias_kind":"arxiv_version","alias_value":"2605.08738v2","created_at":"2026-05-20T00:04:34Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.08738","created_at":"2026-05-20T00:04:34Z"},{"alias_kind":"pith_short_12","alias_value":"B7XDWRI47A3Y","created_at":"2026-05-20T00:04:34Z"},{"alias_kind":"pith_short_16","alias_value":"B7XDWRI47A3YI3JU","created_at":"2026-05-20T00:04:34Z"},{"alias_kind":"pith_short_8","alias_value":"B7XDWRI4","created_at":"2026-05-20T00:04:34Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:B7XDWRI47A3YI3JUJJVZDT5III","target":"record","payload":{"canonical_record":{"source":{"id":"2605.08738","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T06:50:35Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"d690173b393dfc6bc5926ee52dd327d38501a6e48b0392e1e7e4e6f3152c0c84","abstract_canon_sha256":"dd75df23f929b8bad3a1878c02869cdebdf0033ee3e365b579a310e61361a38d"},"schema_version":"1.0"},"canonical_sha256":"0fee3b451cf837846d344a6b91cfa8423e2a12641676538d0577db1ea14f41df","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:04:34.993671Z","signature_b64":"7SbEKK+2iIw8RwAgmTqHiPIs6CYoF/MiXBRu9gbnyKQjXUM6J+hOz0lbUmfaFVLjHTuBp7Ypp1RL3MzxW8wBDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0fee3b451cf837846d344a6b91cfa8423e2a12641676538d0577db1ea14f41df","last_reissued_at":"2026-05-20T00:04:34.992568Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:04:34.992568Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.08738","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":"2026-05-20T00:04:34Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"lbVhJN4DiUB2h6HutSl1HhYE0kYZ6W0c5+4YGIm1CVMYVvse6V0azpGZaVDv0z+gTx8jTVdpgMVsgpt+vDmpCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-28T14:41:11.005807Z"},"content_sha256":"2b1b9c05f6bfd8d697e4f15a739562bcc8e6a7dad5769350e4ed1d45f98bb3b7","schema_version":"1.0","event_id":"sha256:2b1b9c05f6bfd8d697e4f15a739562bcc8e6a7dad5769350e4ed1d45f98bb3b7"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:B7XDWRI47A3YI3JUJJVZDT5III","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Pruning a pretrained large MoE consistently outperforms training the smaller target architecture from scratch under the same training budget.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Bo Zheng, Dayiheng Liu, Liangyu Wang, Rui Men, Shengkun Tang, Siqi Zhang, Xiulong Yuan, Zekun Wang, Zhiqiang Shen, Zihan Qiu","submitted_at":"2026-05-09T06:50:35Z","abstract_excerpt":"Structured pruning and knowledge distillation (KD) are typical techniques for compressing large language models, but it remains unclear how they should be applied at pretraining scale, especially to recent mixture-of-experts (MoE) models. In this work, we systematically study MoE compression in large-scale pretraining, focusing on three key questions: whether pruning provides a better initialization than training from scratch, how expert compression choices affect the final model after continued training, and which training strategy is most effective. We have the following findings: First, acr"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Across depth, width, and expert compression, pruning a pretrained MoE consistently outperforms training the target architecture from scratch under the same training budget; progressive pruning schedules outperform one-shot compression.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the observed convergence of different one-shot expert compression methods after large-scale continual pretraining will generalize beyond the specific Qwen3-Next-80A3B model and the chosen downstream benchmarks.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Pruning pretrained MoE models outperforms training from scratch, different compression methods converge after continued pretraining, and combining KD with language modeling loss plus progressive schedules yields a competitive 23A2B model from Qwen3-Next-80A3B.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Pruning a pretrained large MoE consistently outperforms training the smaller target architecture from scratch under the same training budget.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e7244562dbb9ff2219d209b6c8a080dc63e1c8c334cb18a77d0a968b05b84678"},"source":{"id":"2605.08738","kind":"arxiv","version":2},"verdict":{"id":"e64e0df5-d330-4d2c-8687-08baf1faf45e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T03:24:19.975833Z","strongest_claim":"Across depth, width, and expert compression, pruning a pretrained MoE consistently outperforms training the target architecture from scratch under the same training budget; progressive pruning schedules outperform one-shot compression.","one_line_summary":"Pruning pretrained MoE models outperforms training from scratch, different compression methods converge after continued pretraining, and combining KD with language modeling loss plus progressive schedules yields a competitive 23A2B model from Qwen3-Next-80A3B.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the observed convergence of different one-shot expert compression methods after large-scale continual pretraining will generalize beyond the specific Qwen3-Next-80A3B model and the chosen downstream benchmarks.","pith_extraction_headline":"Pruning a pretrained large MoE consistently outperforms training the smaller target architecture from scratch under the same training budget."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.08738/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T22:35:15.381812Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T14:31:17.561569Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T10:50:48.433480Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"1b83155b8377816cc4861576b945f25ac7ff4ff05efd011084251d8a08e62411"},"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":"e64e0df5-d330-4d2c-8687-08baf1faf45e"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:04:34Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"j/4K2rfPgCDybliLehfhLqavw6KMH5/BBVi40xgw5JAdavVC+SZq9B2IAxlyUnCuPUKKmV5htxVPM+bSRdeICA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-28T14:41:11.006307Z"},"content_sha256":"95a2ca68b413942d70a0742559db48f71ce0efe36410516d0fddd436950f8810","schema_version":"1.0","event_id":"sha256:95a2ca68b413942d70a0742559db48f71ce0efe36410516d0fddd436950f8810"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/B7XDWRI47A3YI3JUJJVZDT5III/bundle.json","state_url":"https://pith.science/pith/B7XDWRI47A3YI3JUJJVZDT5III/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/B7XDWRI47A3YI3JUJJVZDT5III/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":"2026-05-28T14:41:11Z","links":{"resolver":"https://pith.science/pith/B7XDWRI47A3YI3JUJJVZDT5III","bundle":"https://pith.science/pith/B7XDWRI47A3YI3JUJJVZDT5III/bundle.json","state":"https://pith.science/pith/B7XDWRI47A3YI3JUJJVZDT5III/state.json","well_known_bundle":"https://pith.science/.well-known/pith/B7XDWRI47A3YI3JUJJVZDT5III/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:B7XDWRI47A3YI3JUJJVZDT5III","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":"dd75df23f929b8bad3a1878c02869cdebdf0033ee3e365b579a310e61361a38d","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T06:50:35Z","title_canon_sha256":"d690173b393dfc6bc5926ee52dd327d38501a6e48b0392e1e7e4e6f3152c0c84"},"schema_version":"1.0","source":{"id":"2605.08738","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.08738","created_at":"2026-05-20T00:04:34Z"},{"alias_kind":"arxiv_version","alias_value":"2605.08738v2","created_at":"2026-05-20T00:04:34Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.08738","created_at":"2026-05-20T00:04:34Z"},{"alias_kind":"pith_short_12","alias_value":"B7XDWRI47A3Y","created_at":"2026-05-20T00:04:34Z"},{"alias_kind":"pith_short_16","alias_value":"B7XDWRI47A3YI3JU","created_at":"2026-05-20T00:04:34Z"},{"alias_kind":"pith_short_8","alias_value":"B7XDWRI4","created_at":"2026-05-20T00:04:34Z"}],"graph_snapshots":[{"event_id":"sha256:95a2ca68b413942d70a0742559db48f71ce0efe36410516d0fddd436950f8810","target":"graph","created_at":"2026-05-20T00:04:34Z","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":"Across depth, width, and expert compression, pruning a pretrained MoE consistently outperforms training the target architecture from scratch under the same training budget; progressive pruning schedules outperform one-shot compression."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the observed convergence of different one-shot expert compression methods after large-scale continual pretraining will generalize beyond the specific Qwen3-Next-80A3B model and the chosen downstream benchmarks."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Pruning pretrained MoE models outperforms training from scratch, different compression methods converge after continued pretraining, and combining KD with language modeling loss plus progressive schedules yields a competitive 23A2B model from Qwen3-Next-80A3B."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Pruning a pretrained large MoE consistently outperforms training the smaller target architecture from scratch under the same training budget."}],"snapshot_sha256":"e7244562dbb9ff2219d209b6c8a080dc63e1c8c334cb18a77d0a968b05b84678"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T22:35:15.381812Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T14:31:17.561569Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T10:50:48.433480Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.08738/integrity.json","findings":[],"snapshot_sha256":"1b83155b8377816cc4861576b945f25ac7ff4ff05efd011084251d8a08e62411","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Structured pruning and knowledge distillation (KD) are typical techniques for compressing large language models, but it remains unclear how they should be applied at pretraining scale, especially to recent mixture-of-experts (MoE) models. In this work, we systematically study MoE compression in large-scale pretraining, focusing on three key questions: whether pruning provides a better initialization than training from scratch, how expert compression choices affect the final model after continued training, and which training strategy is most effective. We have the following findings: First, acr","authors_text":"Bo Zheng, Dayiheng Liu, Liangyu Wang, Rui Men, Shengkun Tang, Siqi Zhang, Xiulong Yuan, Zekun Wang, Zhiqiang Shen, Zihan Qiu","cross_cats":["cs.AI","cs.CL"],"headline":"Pruning a pretrained large MoE consistently outperforms training the smaller target architecture from scratch under the same training budget.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T06:50:35Z","title":"SlimQwen: Exploring the Pruning and Distillation in Large MoE Model Pre-training"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.08738","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-12T03:24:19.975833Z","id":"e64e0df5-d330-4d2c-8687-08baf1faf45e","model_set":{"reader":"grok-4.3"},"one_line_summary":"Pruning pretrained MoE models outperforms training from scratch, different compression methods converge after continued pretraining, and combining KD with language modeling loss plus progressive schedules yields a competitive 23A2B model from Qwen3-Next-80A3B.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Pruning a pretrained large MoE consistently outperforms training the smaller target architecture from scratch under the same training budget.","strongest_claim":"Across depth, width, and expert compression, pruning a pretrained MoE consistently outperforms training the target architecture from scratch under the same training budget; progressive pruning schedules outperform one-shot compression.","weakest_assumption":"That the observed convergence of different one-shot expert compression methods after large-scale continual pretraining will generalize beyond the specific Qwen3-Next-80A3B model and the chosen downstream benchmarks."}},"verdict_id":"e64e0df5-d330-4d2c-8687-08baf1faf45e"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:2b1b9c05f6bfd8d697e4f15a739562bcc8e6a7dad5769350e4ed1d45f98bb3b7","target":"record","created_at":"2026-05-20T00:04:34Z","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":"dd75df23f929b8bad3a1878c02869cdebdf0033ee3e365b579a310e61361a38d","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-09T06:50:35Z","title_canon_sha256":"d690173b393dfc6bc5926ee52dd327d38501a6e48b0392e1e7e4e6f3152c0c84"},"schema_version":"1.0","source":{"id":"2605.08738","kind":"arxiv","version":2}},"canonical_sha256":"0fee3b451cf837846d344a6b91cfa8423e2a12641676538d0577db1ea14f41df","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"0fee3b451cf837846d344a6b91cfa8423e2a12641676538d0577db1ea14f41df","first_computed_at":"2026-05-20T00:04:34.992568Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:04:34.992568Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"7SbEKK+2iIw8RwAgmTqHiPIs6CYoF/MiXBRu9gbnyKQjXUM6J+hOz0lbUmfaFVLjHTuBp7Ypp1RL3MzxW8wBDQ==","signature_status":"signed_v1","signed_at":"2026-05-20T00:04:34.993671Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.08738","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:2b1b9c05f6bfd8d697e4f15a739562bcc8e6a7dad5769350e4ed1d45f98bb3b7","sha256:95a2ca68b413942d70a0742559db48f71ce0efe36410516d0fddd436950f8810"],"state_sha256":"6d89ea843ee4ecb541e3d8ee087686ccb59a8f3c5e85f3bee689790ba7dc6072"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"FG/jZGdqHhdGI8jF0P9vglO3g6HTBqhRPBuMCz5DHBZQ9wk6R1DaPDa+eqVQ4g9g0Ywd+O1a2bfE+gvj1DCiAw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-28T14:41:11.008734Z","bundle_sha256":"16d4357e85f78805bf2dafc330935af7e14444410c9281d578a4424a21eb6e74"}}