{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:FQXNA77ORWYDEG4KRNLSNDLPM7","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":"4e867976f7f559626b33bccc4a190943e4772af7e96f6da5a2c6426838d31d67","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-16T09:12:52Z","title_canon_sha256":"fa3dfc81e90801d2508f65fa03902b664d983b73726720d46dd5570ad6ea5c86"},"schema_version":"1.0","source":{"id":"2605.16893","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.16893","created_at":"2026-05-20T00:03:28Z"},{"alias_kind":"arxiv_version","alias_value":"2605.16893v1","created_at":"2026-05-20T00:03:28Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.16893","created_at":"2026-05-20T00:03:28Z"},{"alias_kind":"pith_short_12","alias_value":"FQXNA77ORWYD","created_at":"2026-05-20T00:03:28Z"},{"alias_kind":"pith_short_16","alias_value":"FQXNA77ORWYDEG4K","created_at":"2026-05-20T00:03:28Z"},{"alias_kind":"pith_short_8","alias_value":"FQXNA77O","created_at":"2026-05-20T00:03:28Z"}],"graph_snapshots":[{"event_id":"sha256:3242aa4cdd2c5fc0f41fea6886389a5ae74c70c9cf36c6ccfdffee5bc1acf5de","target":"graph","created_at":"2026-05-20T00:03:28Z","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":"NGM improves average performance by 0.5 to 1.2 points, with particularly clear gains on code generation and knowledge-intensive tasks (e.g., +3.0 on LiveCodeBench and +3.03 on GPQA for Qwen3-14B)."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That directly averaging pretrained token embeddings produces useful n-gram representations that the cosine-gated injector can meaningfully modulate without introducing noise or requiring any learned parameters or additional training."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"NGM is a plug-and-play n-gram memory module that encodes n-grams from pretrained embeddings and gates their injection to improve LLM performance by 0.5-1.2 points on average across eight benchmarks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Averaging pretrained token embeddings creates n-gram representations that a cosine-gated injector adds to LLMs without training or extra parameters."}],"snapshot_sha256":"6baecffb6c96d1935dbe25ba7f30ff9e38f2ccd80f94d5b2123a7b3572e0329f"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"380d7b11e8eb1d92f2475716a211ee9061a2ef1158a5922871de50de6256fd8a"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T21:01:19.186204Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"cited_work_retraction","ran_at":"2026-05-19T20:52:41.515826Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T20:51:11.023864Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T18:41:56.281729Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:26.360270Z","status":"skipped","version":"1.0.0"}],"endpoint":"/pith/2605.16893/integrity.json","findings":[],"snapshot_sha256":"89957008473ad97c4417bfbd998e0cd5bb504fc49e8edf985592d4c361433c6e","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Recent studies introduce conditional memory modules that decouple knowledge storage from neural computation, enabling more direct knowledge access. Compared to MoE, which relies on dynamic computation paths, explicit lookup provides a more efficient knowledge retrieval mechanism. However, these approaches still depend on learned memory embeddings, requiring additional training and limiting flexibility. To address this, we propose N-gram Memory (NGM), a training-free, plug-and-play module composed of a Causal N-Gram Encoder and a Cosine-Gated Memory Injector. The Causal N-Gram Encoder directly ","authors_text":"Caifeng Shan, Chenyang Si, Wenhui Dong, Yuwen Qu","cross_cats":[],"headline":"Averaging pretrained token embeddings creates n-gram representations that a cosine-gated injector adds to LLMs without training or extra parameters.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-16T09:12:52Z","title":"NGM: A Plug-and-Play Training-Free Memory Module for LLMs"},"references":{"count":44,"internal_anchors":13,"resolved_work":44,"sample":[{"cited_arxiv_id":"2511.21631","doi":"","is_internal_anchor":true,"ref_index":1,"title":"Qwen3-VL Technical Report","work_id":"1fe243aa-e3c0-4da6-b391-4cbcfc88d5c0","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Enriching word vec- tors with subword information","work_id":"bed31a61-1767-406d-bf52-41211a3ba185","year":2017},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Improving language models by retrieving from trillions of tokens","work_id":"5fd45170-71b7-4233-a7e7-3c1f3c24ac46","year":2022},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Large language models in machine translation","work_id":"0dde0056-d211-4a75-bc5b-eb34faee6977","year":2007},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Are we on the right way for evaluating large vision- language models? Advances in Neural Information Processing Systems , 37:27056–27087, 2024","work_id":"9a68e153-fb33-47fe-b918-443ec4e86a5a","year":2024}],"snapshot_sha256":"16a8bd943eba5ea19e5ca0767d5dbaef6e5182dd7e8a5f7cfd8a373eae45ca4e"},"source":{"id":"2605.16893","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-19T20:44:03.921359Z","id":"653f6d5c-f8ff-469d-996e-09835f4516fc","model_set":{"reader":"grok-4.3"},"one_line_summary":"NGM is a plug-and-play n-gram memory module that encodes n-grams from pretrained embeddings and gates their injection to improve LLM performance by 0.5-1.2 points on average across eight benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Averaging pretrained token embeddings creates n-gram representations that a cosine-gated injector adds to LLMs without training or extra parameters.","strongest_claim":"NGM improves average performance by 0.5 to 1.2 points, with particularly clear gains on code generation and knowledge-intensive tasks (e.g., +3.0 on LiveCodeBench and +3.03 on GPQA for Qwen3-14B).","weakest_assumption":"That directly averaging pretrained token embeddings produces useful n-gram representations that the cosine-gated injector can meaningfully modulate without introducing noise or requiring any learned parameters or additional training."}},"verdict_id":"653f6d5c-f8ff-469d-996e-09835f4516fc"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:c36de5d669618e7a9ee54724161fb3ab21f37459f53c05f7f96c8e6ddc19e475","target":"record","created_at":"2026-05-20T00:03:28Z","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":"4e867976f7f559626b33bccc4a190943e4772af7e96f6da5a2c6426838d31d67","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-16T09:12:52Z","title_canon_sha256":"fa3dfc81e90801d2508f65fa03902b664d983b73726720d46dd5570ad6ea5c86"},"schema_version":"1.0","source":{"id":"2605.16893","kind":"arxiv","version":1}},"canonical_sha256":"2c2ed07fee8db0321b8a8b57268d6f67eb7357c81972eaca89e7737a1790fdf9","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2c2ed07fee8db0321b8a8b57268d6f67eb7357c81972eaca89e7737a1790fdf9","first_computed_at":"2026-05-20T00:03:28.715557Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:03:28.715557Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"TYjdGkVBHtRy9zryZxqKRmrQrh93C318sXf3WfPvkx2aDILQgU2dVEskxIdol6lXI4N4nOV5VgeY+W8ETrVFCA==","signature_status":"signed_v1","signed_at":"2026-05-20T00:03:28.716356Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.16893","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:c36de5d669618e7a9ee54724161fb3ab21f37459f53c05f7f96c8e6ddc19e475","sha256:3242aa4cdd2c5fc0f41fea6886389a5ae74c70c9cf36c6ccfdffee5bc1acf5de"],"state_sha256":"b269f095b3eb4ff12eaa72593e325f6f145f81d44759521bf60c89ceb7a94781"}