{"arxiv_id":"2607.11270","incoming":{"papers":2,"sentences":2,"groups":[{"kind":"background","label":"Cites it as background","papers":2,"sentences":2,"quotes":[{"arxiv_id":"2608.22403","title":"LD4WAM: Learning Latent Dynamics from Human Videos for World Action Models","url":"/paper/2608.22403","section":"World Action Models for Robot Manipulation","sentence":"A second group moves the prediction target into richer feature spaces, using semantic feature [ref] or 3D structure information [ref] .","intent_p":0.89,"contests_p":null},{"arxiv_id":"2608.29768","title":"SmoothRL: Online Reinforcement Learning During Asynchronous Execution","url":"/paper/2608.29768","section":"Introduction","sentence":"Recent advances in Vision-Language-Action (VLA) models and World-Action Models (WAMs) have demonstrated remarkable progress toward endowing robots with general manipulation capabilities [ref] .","intent_p":0.98,"contests_p":null}]}],"unlabelled_sentences":0},"outgoing":{"papers":52,"sentences":81,"groups":[{"kind":"extends","label":"Builds on","papers":1,"sentences":1,"quotes":[{"arxiv_id":"2512.08580","title":null,"url":"/paper/2512.08580","section":"Embodied Understanding","sentence":"Building upon the VLM data curation of Lumo-1 [ref] , we further expand the dataset to emphasize learning physical dynamics from large-scale video data and richer spatial understanding.","intent_p":0.95,"contests_p":null}]},{"kind":"method_or_tool","label":"Uses the method or tool of","papers":6,"sentences":8,"quotes":[{"arxiv_id":"2604.11689","title":"LARY: A Latent Action Representation Yielding Benchmark for Generalizable Vision-to-Action Alignment","url":"/paper/2604.11689","section":"Alignment and Scaling Laws [$𝒬5$]","sentence":"Following the protocol of [ref] , we construct an atomic action dataset based on our robot manipulation data.","intent_p":0.99,"contests_p":null},{"arxiv_id":"2505.06111","title":"UniVLA: Learning to Act Anywhere with Task-centric Latent Actions","url":"/paper/2505.06111","section":"Stage 1: Aligning Actions with Latent World Dynamics","sentence":"Visual features are encoded by a causal spatial-temporal transformer [ref] , projected to a VQ codebook and optimized with a feature-level MSE loss for future feature reconstruction.","intent_p":0.95,"contests_p":null},{"arxiv_id":"2407.12772","title":"LMMs-Eval: Reality Check on the Evaluation of Large Multimodal Models","url":"/paper/2407.12772","section":"VLM Evaluation [$𝒬1$]","sentence":"Tab. [ref] summarizes the evaluation results of and competing 4B-scale models on two categories of embodied benchmarks, namely embodied understanding and embodied location, with all tests performed via the lmms-eval [ref] framework.","intent_p":1.0,"contests_p":null},{"arxiv_id":"2512.08580","title":null,"url":"/paper/2512.08580","section":"Generalizable Pick and Place","sentence":"Following Lumo-1 [ref] , we evaluate model instruction following and generalization capabilities on the task of generalizable pick and place with “put A into/to B” instruction, where A is the object and B is the target location.","intent_p":0.96,"contests_p":null},{"arxiv_id":"2507.17141","title":"Towards Human-level Intelligence via Human-like Whole-Body Manipulation","url":"/paper/2507.17141","section":"Task Post-Training [$𝒬3$]","sentence":"All experiments are conducted on the high-performance bimanual mobile manipulation platform Astribot S1 [ref] .","intent_p":0.92,"contests_p":null}]},{"kind":"data_used","label":"Uses data or values from","papers":3,"sentences":3,"quotes":[{"arxiv_id":"2503.06669","title":"AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems","url":"/paper/2503.06669","section":"Stage 1: Aligning Actions with Latent World Dynamics","sentence":"We train our model on a heterogeneous, multi-source corpus comprising self-collected Astribot S1 [ref] manipulation data, cross-embodiment robot datasets spanning diverse platforms - including bimanual humanoids such as AGIBot Genie-1 [ref] , Galaxea R1 Pro [ref] , and RealSource RS-02 [ref] , as well as tabletop dual-arm platforms like ARX, Agile X, and YAM.","intent_p":0.98,"contests_p":null},{"arxiv_id":"2507.17141","title":"Towards Human-level Intelligence via Human-like Whole-Body Manipulation","url":"/paper/2507.17141","section":"Stage 1: Aligning Actions with Latent World Dynamics","sentence":"We train our model on a heterogeneous, multi-source corpus comprising self-collected Astribot S1 [ref] manipulation data, cross-embodiment robot datasets spanning diverse platforms - including bimanual humanoids such as AGIBot Genie-1 [ref] , Galaxea R1 Pro [ref] , and RealSource RS-02 [ref] , as well as tabletop dual-arm platforms like ARX, Agile X, and YAM.","intent_p":0.98,"contests_p":null},{"arxiv_id":"2509.00576","title":"Galaxea Open-World Dataset and G0 Dual-System VLA Model","url":"/paper/2509.00576","section":"Stage 1: Aligning Actions with Latent World Dynamics","sentence":"We train our model on a heterogeneous, multi-source corpus comprising self-collected Astribot S1 [ref] manipulation data, cross-embodiment robot datasets spanning diverse platforms - including bimanual humanoids such as AGIBot Genie-1 [ref] , Galaxea R1 Pro [ref] , and RealSource RS-02 [ref] , as well as tabletop dual-arm platforms like ARX, Agile X, and YAM.","intent_p":0.98,"contests_p":null}]},{"kind":"comparison","label":"Compares with","papers":5,"sentences":11,"quotes":[{"arxiv_id":"2504.16054","title":"$\\pi_{0.5}$: a Vision-Language-Action Model with Open-World Generalization","url":"/paper/2504.16054","section":"Results.","sentence":"Detailed per-task success rates are reported in Fig. [ref], and category‑level comparisons are provided in Fig. [ref]. outperforms both baseline models $π_0.5$ [ref] and Fast-WAM [ref] across all task categories.","intent_p":1.0,"contests_p":null},{"arxiv_id":"2603.16666","title":"Fast-WAM: Do World Action Models Need Test-time Future Imagination?","url":"/paper/2603.16666","section":"Results.","sentence":"Detailed per-task success rates are reported in Fig. [ref], and category‑level comparisons are provided in Fig. [ref]. outperforms both baseline models $π_0.5$ [ref] and Fast-WAM [ref] across all task categories.","intent_p":1.0,"contests_p":null},{"arxiv_id":"2406.10721","title":"RoboPoint: A Vision-Language Model for Spatial Affordance Prediction for Robotics","url":"/paper/2406.10721","section":"VLM Evaluation [$𝒬1$]","sentence":"It achieves the best performance on multiple benchmarks including VSIBench, MindCube [ref] , and RefSpatialBench [ref] ; it also ranks first on the Where2place [ref] task in the embodied location category, establishing its strong competence in embodied perception and spatial cognition.","intent_p":0.86,"contests_p":null},{"arxiv_id":"2602.14979","title":null,"url":"/paper/2602.14979","section":"VLM Evaluation [$𝒬1$]","sentence":"Against other state-of-the-art 4B-scale embodied models (RoboBrain-2.5 4B [ref] and Rynnbrain 4B [ref] ), VLM delivers highly competitive results.","intent_p":0.99,"contests_p":null},{"arxiv_id":"2601.14352","title":null,"url":"/paper/2601.14352","section":"VLM Evaluation [$𝒬1$]","sentence":"Against other state-of-the-art 4B-scale embodied models (RoboBrain-2.5 4B [ref] and Rynnbrain 4B [ref] ), VLM delivers highly competitive results.","intent_p":0.99,"contests_p":null}]},{"kind":"background","label":"Cites as background","papers":19,"sentences":23,"quotes":[{"arxiv_id":"2602.03310","title":null,"url":"/paper/2602.03310","section":"Progressive Action Tokenization and Modality Alignment","sentence":"In practice, actions are typically modeled either as raw continuous signals, or as discretized tokens optimized for high-fidelity reconstruction, evolving from simple binning schemes [ref] to more advanced approaches such as FAST [ref] and VAE-based autoencoders [ref] .","intent_p":0.91,"contests_p":null},{"arxiv_id":"2501.09747","title":"FAST: Efficient Action Tokenization for Vision-Language-Action Models","url":"/paper/2501.09747","section":"Progressive Action Tokenization and Modality Alignment","sentence":"In practice, actions are typically modeled either as raw continuous signals, or as discretized tokens optimized for high-fidelity reconstruction, evolving from simple binning schemes [ref] to more advanced approaches such as FAST [ref] and VAE-based autoencoders [ref] .","intent_p":0.91,"contests_p":null},{"arxiv_id":"2512.04952","title":null,"url":"/paper/2512.04952","section":"Progressive Action Tokenization and Modality Alignment","sentence":"In practice, actions are typically modeled either as raw continuous signals, or as discretized tokens optimized for high-fidelity reconstruction, evolving from simple binning schemes [ref] to more advanced approaches such as FAST [ref] and VAE-based autoencoders [ref] .","intent_p":0.91,"contests_p":null},{"arxiv_id":"2406.09246","title":"OpenVLA: An Open-Source Vision-Language-Action Model","url":"/paper/2406.09246","section":"Progressive Action Tokenization and Modality Alignment","sentence":"In practice, actions are typically modeled either as raw continuous signals, or as discretized tokens optimized for high-fidelity reconstruction, evolving from simple binning schemes [ref] to more advanced approaches such as FAST [ref] and VAE-based autoencoders [ref] .","intent_p":0.91,"contests_p":null},{"arxiv_id":"2508.05635","title":"Genie Envisioner: A Unified World Foundation Platform for Robotic Manipulation","url":"/paper/2508.05635","section":"Related Work","sentence":"One line of work employs video models as predictive representation learners or transferable world priors, typically followed by separate action decoding modules [ref] .","intent_p":0.87,"contests_p":null}]}],"unlabelled_sentences":35},"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/2607.11270"}}