{"arxiv_id":"2512.21711","incoming":{"papers":14,"sentences":23,"groups":[{"kind":"contests","label":"Contests it","papers":1,"sentences":1,"quotes":[{"arxiv_id":"2607.18264","title":"MUX: Continuous Reasoning via Multiplexed Tokens","url":"/paper/2607.18264","section":"Introduction","sentence":"One class of methods relies on temporal backpropagation of trajectory-level losses [ref] , which tend to produce shortcut or uninformative latent tokens [ref] .","intent_p":0.44,"contests_p":0.77}]},{"kind":"method_or_tool","label":"Uses its method or tool","papers":1,"sentences":1,"quotes":[{"arxiv_id":"2602.22766","title":"Imagination Helps Visual Reasoning, But Not Yet in Latent Space","url":"/paper/2602.22766","section":"Dependency Analysis","sentence":"For interventions on $Z$, we follow the protocol of [ref] and explicitly manipulate the reasoning process.","intent_p":1.0,"contests_p":null}]},{"kind":"supports_this_paper","label":"Cites it as support","papers":4,"sentences":4,"quotes":[{"arxiv_id":"2606.03768","title":"HybridThinker: Efficient Chain-of-Thought Reasoning via Compressed Memory and Transient Thought Steps","url":"/paper/2606.03768","section":"Introduction","sentence":"Unfortunately, this premature discarding renders the immediately following step more error-prone: key details such as exact numerical values, strict constraints, and intermediate conclusions may be obscured or lost in the compact representations [ref] .","intent_p":0.97,"contests_p":null},{"arxiv_id":"2606.20075","title":"What Makes Effective Supervision in Latent Chain-of-Thought: An Information-Theoretic Analysis","url":"/paper/2606.20075","section":"Supervision Challenges in Latent CoT","sentence":"In high-dimensional continuous spaces, the model may instead rely on shortcut features or spurious representations that are unrelated to valid reasoning [ref] .","intent_p":0.98,"contests_p":null},{"arxiv_id":"2608.11218","title":"MaSRead: Content-Addressed Reading of Replicated Latent Stores","url":"/paper/2608.11218","section":"Related work","sentence":"Latent tokens transport and scaffold computation rather than perform it [ref] , and multi-hop composition is built by a model's own attention over a sequential chain rather than by colocating independent vectors [ref] .","intent_p":1.0,"contests_p":null},{"arxiv_id":"2604.14889","title":"MemoSight: Unifying Context Compression and Multi Token Prediction for Reasoning Acceleration","url":"/paper/2604.14889","section":"Related Work","sentence":"However, latent reasoning lacks interpretability and is prone to learning dataset artifacts in its reasoning [ref] .","intent_p":0.96,"contests_p":null}]},{"kind":"background","label":"Cites it as background","papers":1,"sentences":1,"quotes":[{"arxiv_id":"2602.22766","title":"Imagination Helps Visual Reasoning, But Not Yet in Latent Space","url":"/paper/2602.22766","section":"Introduction","sentence":"Building on these properties, a series of latent-space visual reasoning methods have been proposed [ref] .","intent_p":0.91,"contests_p":null}]}],"unlabelled_sentences":16},"outgoing":{"papers":0,"sentences":0,"groups":[],"unlabelled_sentences":0},"claims":[],"thresholds":{"intent_p":0.8,"supports_p":0.95,"contests_p":0.7},"method":"Citing sentences come from each paper's LaTeX source; every \\cite marks the sentence and the works it cites. Pith's classifier typed each sentence. A label is shown at intent probability 0.8 or more (0.95 for support), a contest at 0.7 or more. Measured accuracy against two blind model families is at https://pith.science/accuracy. The quoted sentences are the citing authors' own.","held_up":null,"_links":{"html":"https://pith.science/paper/2512.21711"}}