{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:I4DGHULBUTOM3UHCW6HW2J7LII","short_pith_number":"pith:I4DGHULB","schema_version":"1.0","canonical_sha256":"470663d161a4dccdd0e2b78f6d27eb42303a775c9d9a9424ba44203ec470e162","source":{"kind":"arxiv","id":"2103.10385","version":2},"attestation_state":"computed","paper":{"title":"GPT Understands, Too","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Jie Tang, Ming Ding, Xiao Liu, Yanan Zheng, Yujie Qian, Zhengxiao Du, Zhilin Yang","submitted_at":"2021-03-18T17:13:50Z","abstract_excerpt":"Prompting a pretrained language model with natural language patterns has been proved effective for natural language understanding (NLU). However, our preliminary study reveals that manual discrete prompts often lead to unstable performance -- e.g., changing a single word in the prompt might result in substantial performance drop. We propose a novel method P-Tuning that employs trainable continuous prompt embeddings in concatenation with discrete prompts. Empirically, P-Tuning not only stabilizes training by minimizing the gap between various discrete prompts, but also improves performance by a"},"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":"2103.10385","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-03-18T17:13:50Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"fc6b781da8302fbde12766804a6d0e849efd13b802713fdf5c7d46089e73c5f8","abstract_canon_sha256":"18324fdae5d67f1a648e80265b47619493f176ca35731fb66d5533928ac91eb4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:04:40.505972Z","signature_b64":"ROSY44l2emV2KIfh8dWm5SGOJlSGFmbVhkZbgLrFCBhlMtadKDuYItqT0s9p/Pj/sM0eXtrr4bGgpS0Mlk04Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"470663d161a4dccdd0e2b78f6d27eb42303a775c9d9a9424ba44203ec470e162","last_reissued_at":"2026-07-05T07:04:40.505441Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:04:40.505441Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GPT Understands, Too","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Jie Tang, Ming Ding, Xiao Liu, Yanan Zheng, Yujie Qian, Zhengxiao Du, Zhilin Yang","submitted_at":"2021-03-18T17:13:50Z","abstract_excerpt":"Prompting a pretrained language model with natural language patterns has been proved effective for natural language understanding (NLU). However, our preliminary study reveals that manual discrete prompts often lead to unstable performance -- e.g., changing a single word in the prompt might result in substantial performance drop. We propose a novel method P-Tuning that employs trainable continuous prompt embeddings in concatenation with discrete prompts. Empirically, P-Tuning not only stabilizes training by minimizing the gap between various discrete prompts, but also improves performance by a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2103.10385","kind":"arxiv","version":2},"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/2103.10385/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":"2103.10385","created_at":"2026-07-05T07:04:40.505499+00:00"},{"alias_kind":"arxiv_version","alias_value":"2103.10385v2","created_at":"2026-07-05T07:04:40.505499+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2103.10385","created_at":"2026-07-05T07:04:40.505499+00:00"},{"alias_kind":"pith_short_12","alias_value":"I4DGHULBUTOM","created_at":"2026-07-05T07:04:40.505499+00:00"},{"alias_kind":"pith_short_16","alias_value":"I4DGHULBUTOM3UHC","created_at":"2026-07-05T07:04:40.505499+00:00"},{"alias_kind":"pith_short_8","alias_value":"I4DGHULB","created_at":"2026-07-05T07:04:40.505499+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":20,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24841","citing_title":"Matching Tasks to Objectives: Fine-Tuning and Prompt-Tuning Strategies for Encoder-Decoder Pre-trained Language Models","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.31246","citing_title":"BadBone: Backdoor Attacks Against Backbone Models in Visual Prompt Learning","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2211.16327","citing_title":"On the Power of Foundation Models","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2305.09617","citing_title":"Towards Expert-Level Medical Question Answering with Large Language Models","ref_index":76,"is_internal_anchor":false},{"citing_arxiv_id":"2307.06435","citing_title":"A Comprehensive Overview of Large Language Models","ref_index":247,"is_internal_anchor":false},{"citing_arxiv_id":"2305.07922","citing_title":"CodeT5+: Open Code Large Language Models for Code Understanding and Generation","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2104.08773","citing_title":"Cross-Task Generalization via Natural Language Crowdsourcing Instructions","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2309.08532","citing_title":"EvoPrompt: Connecting LLMs with Evolutionary Algorithms Yields Powerful Prompt Optimizers","ref_index":110,"is_internal_anchor":false},{"citing_arxiv_id":"2305.14233","citing_title":"Enhancing Chat Language Models by Scaling High-quality Instructional Conversations","ref_index":165,"is_internal_anchor":false},{"citing_arxiv_id":"2202.12837","citing_title":"Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?","ref_index":119,"is_internal_anchor":false},{"citing_arxiv_id":"2507.03724","citing_title":"MemOS: A Memory OS for AI System","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2309.03409","citing_title":"Large Language Models as Optimizers","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2303.16199","citing_title":"LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2403.14608","citing_title":"Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2104.08691","citing_title":"The Power of Scale for Parameter-Efficient Prompt Tuning","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2502.18864","citing_title":"Towards an AI co-scientist","ref_index":158,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04058","citing_title":"MP-ISMoE: Mixed-Precision Interactive Side Mixture-of-Experts for Efficient Transfer Learning","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05732","citing_title":"Graph Topology Information Enhanced Heterogeneous Graph Representation Learning","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2205.01068","citing_title":"OPT: Open Pre-trained Transformer Language Models","ref_index":133,"is_internal_anchor":false},{"citing_arxiv_id":"2106.09685","citing_title":"LoRA: Low-Rank Adaptation of Large Language Models","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/I4DGHULBUTOM3UHCW6HW2J7LII","json":"https://pith.science/pith/I4DGHULBUTOM3UHCW6HW2J7LII.json","graph_json":"https://pith.science/api/pith-number/I4DGHULBUTOM3UHCW6HW2J7LII/graph.json","events_json":"https://pith.science/api/pith-number/I4DGHULBUTOM3UHCW6HW2J7LII/events.json","paper":"https://pith.science/paper/I4DGHULB"},"agent_actions":{"view_html":"https://pith.science/pith/I4DGHULBUTOM3UHCW6HW2J7LII","download_json":"https://pith.science/pith/I4DGHULBUTOM3UHCW6HW2J7LII.json","view_paper":"https://pith.science/paper/I4DGHULB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2103.10385&json=true","fetch_graph":"https://pith.science/api/pith-number/I4DGHULBUTOM3UHCW6HW2J7LII/graph.json","fetch_events":"https://pith.science/api/pith-number/I4DGHULBUTOM3UHCW6HW2J7LII/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/I4DGHULBUTOM3UHCW6HW2J7LII/action/timestamp_anchor","attest_storage":"https://pith.science/pith/I4DGHULBUTOM3UHCW6HW2J7LII/action/storage_attestation","attest_author":"https://pith.science/pith/I4DGHULBUTOM3UHCW6HW2J7LII/action/author_attestation","sign_citation":"https://pith.science/pith/I4DGHULBUTOM3UHCW6HW2J7LII/action/citation_signature","submit_replication":"https://pith.science/pith/I4DGHULBUTOM3UHCW6HW2J7LII/action/replication_record"}},"created_at":"2026-07-05T07:04:40.505499+00:00","updated_at":"2026-07-05T07:04:40.505499+00:00"}