{"arxiv_id":"2608.08023","incoming":{"papers":0,"sentences":0,"groups":[],"unlabelled_sentences":0},"outgoing":{"papers":24,"sentences":40,"groups":[{"kind":"method_or_tool","label":"Uses the method or tool of","papers":1,"sentences":1,"quotes":[{"arxiv_id":"1805.07836","title":"Generalized Cross Entropy Loss for Training Deep Neural Networks with Noisy Labels","url":"/paper/1805.07836","section":"Probing Experiments: Mechanism of 4D Awareness","sentence":"We report KL divergence [ref] and cross-entropy (CE) [ref] between the probe prediction and the Trace Anything target distribution.","intent_p":0.99,"contests_p":null}]},{"kind":"data_used","label":"Uses data or values from","papers":3,"sentences":3,"quotes":[{"arxiv_id":"2506.18088","title":"RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulation","url":"/paper/2506.18088","section":"Introduction","sentence":"We conduct extensive in-distribution experiments on LIBERO [ref] and RoboTwin 2.0 [ref] , as well as out-of-distribution evaluations on LIBERO-Plus [ref] and RoboTwin Clean2Rand [ref] .","intent_p":0.87,"contests_p":null},{"arxiv_id":"2510.13626","title":"LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Models","url":"/paper/2510.13626","section":"Introduction","sentence":"We conduct extensive in-distribution experiments on LIBERO [ref] and RoboTwin 2.0 [ref] , as well as out-of-distribution evaluations on LIBERO-Plus [ref] and RoboTwin Clean2Rand [ref] .","intent_p":0.87,"contests_p":null},{"arxiv_id":"2606.17846","title":"Qwen-RobotManip Technical Report: Alignment Unlocks Scale for Robotic Manipulation Foundation Models","url":"/paper/2606.17846","section":"Introduction","sentence":"We conduct extensive in-distribution experiments on LIBERO [ref] and RoboTwin 2.0 [ref] , as well as out-of-distribution evaluations on LIBERO-Plus [ref] and RoboTwin Clean2Rand [ref] .","intent_p":0.87,"contests_p":null}]},{"kind":"supports_this_paper","label":"Cites as support","papers":2,"sentences":2,"quotes":[{"arxiv_id":"2410.06940","title":"Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Think","url":"/paper/2410.06940","section":"Ablation Study","sentence":"Moreover, we evaluate different alignment layers to assess their impact on performance, motivated by prior findings that intermediate layers are most effective for representation alignment [ref] .","intent_p":0.99,"contests_p":null},{"arxiv_id":"2510.12276","title":null,"url":"/paper/2510.12276","section":"Ablation Study","sentence":"Moreover, we evaluate different alignment layers to assess their impact on performance, motivated by prior findings that intermediate layers are most effective for representation alignment [ref] .","intent_p":0.99,"contests_p":null}]},{"kind":"background","label":"Cites as background","papers":11,"sentences":11,"quotes":[{"arxiv_id":"2604.26694","title":"Unified 4D World Action Modeling from Video Priors with Asynchronous Denoising","url":"/paper/2604.26694","section":"Spatiotemporal Understanding for WAMs","sentence":"WAM4D transfers geometric knowledge into transformers through spatial tokens [ref] , while X-WAM introduces RGB-D video prediction and 3D reconstruction to provide explicit spatial supervision [ref] . $μ_0$ learns an interactive trace world model that predicts future 3D interaction traces and contact regions [ref] .","intent_p":0.94,"contests_p":null},{"arxiv_id":"2606.14048","title":"WAM4D: Fast 4D World Action Model via Spatial Register Tokens","url":"/paper/2606.14048","section":"Spatiotemporal Understanding for WAMs","sentence":"WAM4D transfers geometric knowledge into transformers through spatial tokens [ref] , while X-WAM introduces RGB-D video prediction and 3D reconstruction to provide explicit spatial supervision [ref] . $μ_0$ learns an interactive trace world model that predicts future 3D interaction traces and contact regions [ref] .","intent_p":0.94,"contests_p":null},{"arxiv_id":"2606.13769","title":"$\\mu_0$: A Scalable 3D Interaction-Trace World Model","url":"/paper/2606.13769","section":"Spatiotemporal Understanding for WAMs","sentence":"WAM4D transfers geometric knowledge into transformers through spatial tokens [ref] , while X-WAM introduces RGB-D video prediction and 3D reconstruction to provide explicit spatial supervision [ref] . $μ_0$ learns an interactive trace world model that predicts future 3D interaction traces and contact regions [ref] .","intent_p":0.94,"contests_p":null},{"arxiv_id":"2504.20995","title":"TesserAct: Learning 4D Embodied World Models","url":"/paper/2504.20995","section":"Spatiotemporal Understanding for WAMs","sentence":"TesserAct [ref] subsequently extends world model outputs from 2D videos to 4D scene dynamics by jointly generating RGB, depth, and surface-normal sequences for 4D reconstruction.","intent_p":0.81,"contests_p":null},{"arxiv_id":"2601.03782","title":null,"url":"/paper/2601.03782","section":"Spatiotemporal Understanding for WAMs","sentence":"PointWorld [ref] and Kinema4D [ref] both convert robot actions into spatially grounded point trajectories.","intent_p":0.91,"contests_p":null}]}],"unlabelled_sentences":23},"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/2608.08023"}}