{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BTDHHYDB4A6GCTX3J7W5ZFC4RF","short_pith_number":"pith:BTDHHYDB","schema_version":"1.0","canonical_sha256":"0cc673e061e03c614efb4feddc945c89760088d0037839332a0f25e0af2c84e5","source":{"kind":"arxiv","id":"2505.15778","version":1},"attestation_state":"computed","paper":{"title":"Soft Thinking: Unlocking the Reasoning Potential of LLMs in Continuous Concept Space","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ao Shen, Chenyang Zhao, Shuohang Wang, Weixiang Yan, Xin Eric Wang, Xuehai He, Yelong Shen, Zhen Zhang","submitted_at":"2025-05-21T17:29:15Z","abstract_excerpt":"Human cognition typically involves thinking through abstract, fluid concepts rather than strictly using discrete linguistic tokens. Current reasoning models, however, are constrained to reasoning within the boundaries of human language, processing discrete token embeddings that represent fixed points in the semantic space. This discrete constraint restricts the expressive power and upper potential of such reasoning models, often causing incomplete exploration of reasoning paths, as standard Chain-of-Thought (CoT) methods rely on sampling one token per step. In this work, we introduce Soft Thin"},"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":"2505.15778","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-21T17:29:15Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"d76cff0bdb5759d456741aaf4b68894326340cee558afdb4a79b4435e5d607b6","abstract_canon_sha256":"2b424e90d04c23c16f4de0ef09002f8fbeb5270987783237e86e901c1873cb21"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:06:57.829847Z","signature_b64":"jDBorusqFyoVDSxhQ+CElQTNiHsJ1e82p3R/q8RBmbZegdzoAbMXrBzfzCwcgjBvu3b+x2nn6R58tEvDtINuDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0cc673e061e03c614efb4feddc945c89760088d0037839332a0f25e0af2c84e5","last_reissued_at":"2026-07-05T11:06:57.829390Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:06:57.829390Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Soft Thinking: Unlocking the Reasoning Potential of LLMs in Continuous Concept Space","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ao Shen, Chenyang Zhao, Shuohang Wang, Weixiang Yan, Xin Eric Wang, Xuehai He, Yelong Shen, Zhen Zhang","submitted_at":"2025-05-21T17:29:15Z","abstract_excerpt":"Human cognition typically involves thinking through abstract, fluid concepts rather than strictly using discrete linguistic tokens. Current reasoning models, however, are constrained to reasoning within the boundaries of human language, processing discrete token embeddings that represent fixed points in the semantic space. This discrete constraint restricts the expressive power and upper potential of such reasoning models, often causing incomplete exploration of reasoning paths, as standard Chain-of-Thought (CoT) methods rely on sampling one token per step. In this work, we introduce Soft Thin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.15778","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/2505.15778/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":"2505.15778","created_at":"2026-07-05T11:06:57.829450+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.15778v1","created_at":"2026-07-05T11:06:57.829450+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.15778","created_at":"2026-07-05T11:06:57.829450+00:00"},{"alias_kind":"pith_short_12","alias_value":"BTDHHYDB4A6G","created_at":"2026-07-05T11:06:57.829450+00:00"},{"alias_kind":"pith_short_16","alias_value":"BTDHHYDB4A6GCTX3","created_at":"2026-07-05T11:06:57.829450+00:00"},{"alias_kind":"pith_short_8","alias_value":"BTDHHYDB","created_at":"2026-07-05T11:06:57.829450+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":23,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25354","citing_title":"Efficient and Trainable Language Model Test-Time Scaling via Local Branch Routing","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2606.13106","citing_title":"Demystifying Hidden-State Recurrence: Switchable Latent Reasoning with On-Policy Reinforcement Learning","ref_index":66,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12689","citing_title":"Observable Patterns Are Not Explanations: A Causal-Geometric Analysis of Latent Reasoning Models","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10768","citing_title":"N-GRPO: Embedding-Level Neighbor Mixing for Enhanced Policy Optimization","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00341","citing_title":"DiscoLoop: Looping Discrete Embeddings and Continuous Hidden States for Multi-hop Reasoning","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02248","citing_title":"Geometric Latent Reasoning Induces Shorter Generations in LLMs","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.25354","citing_title":"Efficient and Trainable Language Model Test-Time Scaling via Local Branch Routing","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29928","citing_title":"Latent-CURE for Breast Cancer Diagnosis","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01168","citing_title":"Thinking Economically: A Hierarchical Framework for Adaptive-Complexity Reasoning in LLMs","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10184","citing_title":"Dropout-GRPO: Variational Stochasticity for Continuous Latent Reasoning","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19376","citing_title":"Generative Recursive Reasoning","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17187","citing_title":"PluRule: A Benchmark for Moderating Pluralistic Communities on Social Media","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19376","citing_title":"Generative Recursive Reasoning","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16961","citing_title":"Latent Action Control for Reasoning-Guided Unified Image Generation","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2509.02547","citing_title":"The Landscape of Agentic Reinforcement Learning for LLMs: A Survey","ref_index":205,"is_internal_anchor":false},{"citing_arxiv_id":"2512.12623","citing_title":"Reasoning Within the Mind: Dynamic Multimodal Interleaving in Latent Space","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2602.09850","citing_title":"Towards Explainable Industrial Anomaly Detection via Knowledge-Guided Latent Reasoning","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2603.12529","citing_title":"TERMINATOR: Learning Optimal Exit Points for Early Stopping in Chain-of-Thought Reasoning","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03679","citing_title":"LightThinker++: From Reasoning Compression to Memory Management","ref_index":79,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11856","citing_title":"UniVLR: Unifying Text and Vision in Visual Latent Reasoning for Multimodal LLMs","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22709","citing_title":"Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thought","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08299","citing_title":"SeLaR: Selective Latent Reasoning in Large Language Models","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17892","citing_title":"LEPO: Latent Reasoning Policy Optimization for Large Language Models","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BTDHHYDB4A6GCTX3J7W5ZFC4RF","json":"https://pith.science/pith/BTDHHYDB4A6GCTX3J7W5ZFC4RF.json","graph_json":"https://pith.science/api/pith-number/BTDHHYDB4A6GCTX3J7W5ZFC4RF/graph.json","events_json":"https://pith.science/api/pith-number/BTDHHYDB4A6GCTX3J7W5ZFC4RF/events.json","paper":"https://pith.science/paper/BTDHHYDB"},"agent_actions":{"view_html":"https://pith.science/pith/BTDHHYDB4A6GCTX3J7W5ZFC4RF","download_json":"https://pith.science/pith/BTDHHYDB4A6GCTX3J7W5ZFC4RF.json","view_paper":"https://pith.science/paper/BTDHHYDB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.15778&json=true","fetch_graph":"https://pith.science/api/pith-number/BTDHHYDB4A6GCTX3J7W5ZFC4RF/graph.json","fetch_events":"https://pith.science/api/pith-number/BTDHHYDB4A6GCTX3J7W5ZFC4RF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BTDHHYDB4A6GCTX3J7W5ZFC4RF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BTDHHYDB4A6GCTX3J7W5ZFC4RF/action/storage_attestation","attest_author":"https://pith.science/pith/BTDHHYDB4A6GCTX3J7W5ZFC4RF/action/author_attestation","sign_citation":"https://pith.science/pith/BTDHHYDB4A6GCTX3J7W5ZFC4RF/action/citation_signature","submit_replication":"https://pith.science/pith/BTDHHYDB4A6GCTX3J7W5ZFC4RF/action/replication_record"}},"created_at":"2026-07-05T11:06:57.829450+00:00","updated_at":"2026-07-05T11:06:57.829450+00:00"}