{"total":15,"items":[{"citing_arxiv_id":"2606.30318","ref_index":65,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Chronos: A Physics-Informed Full-History Framework for Non-Markovian Long-Horizon Manipulation","primary_cat":"cs.RO","submitted_at":"2026-06-29T14:00:17+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Chronos elevates full observation history to the policy's latent state via selective SSM tokens and a Schrödinger-inspired acceleration bridge, achieving large gains on memory-dependent robot tasks with fewer parameters.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.26361","ref_index":15,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Does Aurora Encode Atmospheric Structure? Latent Regime Analysis and Attribution","primary_cat":"cs.LG","submitted_at":"2026-06-24T20:18:12+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Aurora's latent space is organized by seasonal cycles with evidence of encoding 3D vertical atmospheric structure for storms, confirmed by perturbation experiments.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.12215","ref_index":15,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"MLT-Dedup: Efficient Large-Scale Online Video Deduplication via Multi-Level Representations and Spatial-Temporal Matching","primary_cat":"cs.CV","submitted_at":"2026-06-10T15:29:45+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"MLT-Dedup achieves 91% reduction in online video repetition rates at 90% precision and 5x indexing capacity using multi-level representations and differential feature-enhanced similarity on a real-world platform.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.04857","ref_index":8,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Rethinking Incompleteness: Formalizing Protocol Divergence and Train-Once Learning for Robust IMVC","primary_cat":"cs.LG","submitted_at":"2026-06-03T13:24:09+00:00","verdict":"ACCEPT","verdict_confidence":"HIGH","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Reconstruction-based IMVC is structurally untrainable when complete-sample proportion falls near zero; CRAFT escapes that bound via per-sample attention-masked fusion trained once on complete data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.01495","ref_index":6,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"CART: Context-Anchored Recurrent Transformer -- A Parameter-Efficient Architecture with Learned Stability","primary_cat":"cs.LG","submitted_at":"2026-05-31T23:26:27+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"CART is a recurrent transformer with shared core, frozen prelude KV tensors, and LTI stability gate that fails to beat dense baselines at parameter parity across tested widths.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.26540","ref_index":38,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Domain-Gated Latent Diffusion: Generative Inverse Design of HMX-Class Energetic Materials with First-Principles Validation","primary_cat":"physics.chem-ph","submitted_at":"2026-05-26T04:43:45+00:00","verdict":null,"verdict_confidence":null,"novelty_score":null,"formal_verification":null,"one_line_summary":null,"context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.09989","ref_index":49,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"StereoPolicy: Improving Robotic Manipulation Policies via Stereo Perception","primary_cat":"cs.RO","submitted_at":"2026-05-11T05:06:12+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"StereoPolicy fuses left-right image features via cross-attention to deliver consistent gains over RGB, RGB-D, point cloud, and multi-view baselines in simulation and real-robot manipulation tasks.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"stereo baseline-to-distance ratio, to derive practical guidelines for configuring stereo setups. Model Design.We comprehensively study stereo encoder design, comparing different backbone choices and revealing an effective architecture for robust stereo-based manipulation. 2 Related Work 2D and 3D Visual RepresentationsVisual encoders have played a central role in both perception and control, with 2D vision backbones such as DINOv3 [49], CLIP [50], and ViT [15] demon- strating remarkable generalization across diverse visual domains. In contrast, 3D vision encoders explicitly model geometric information using volumetric, point-cloud [51-55], or multi-view rep- resentations [56], yet their performance often suffers from the scarcity of large-scale annotated 3D 2 Figure 2:StereoPolicy Pipeline."},{"citing_arxiv_id":"2605.02300","ref_index":264,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"A Meta Reinforcement Learning Approach to Goals-Based Wealth Management","primary_cat":"cs.LG","submitted_at":"2026-05-04T07:48:02+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"MetaRL pre-trained on GBWM problems delivers near-optimal dynamic strategies in 0.01s achieving 97.8% of DP optimal utility and handles larger problems where DP fails.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.23750","ref_index":18,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"The Override Gap: A Magnitude Account of Knowledge Conflict Failure in Hypernetwork-Based Instant LLM Adaptation","primary_cat":"cs.LG","submitted_at":"2026-04-26T14:59:14+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Knowledge conflicts in hypernetwork LLM adaptation stem from constant adapter margins losing to frequency-dependent pretrained margins; selective layer boosting and conflict-aware triggering raise deep-conflict accuracy to 71-72.5% on Gemma-2B and Mistral-7B.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"hypernetworkH θ produces the per-layer adapter matrices {Al, Bl}L l=1 =H θ(d),(2) whereLis the number of target transformer layers. The base model's effective weights during inference become fWl =W l + (α/r)B lAl. Producing the adapter takes a single forward pass of Hθ, which is typically less than one second on a modern GPU. Doc-to-LoRA [3] implementsH θ as a Perceiver-style cross-attention module [18] that reads per- layer activations from a frozen context encoder and outputs rank-8 adapters for the feed forward down-projection layers. SHINE [24] reuses the backbone language model as its own context encoder and employs alternating row-column attention for cross-layer communication during generation. Training proceeds by minimizing a reconstruction loss on a corpus of documents: givend, the"},{"citing_arxiv_id":"2604.22442","ref_index":17,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"HubRouter: A Pluggable Sub-Quadratic Routing Primitive for Hybrid Sequence Models","primary_cat":"cs.LG","submitted_at":"2026-04-24T10:59:30+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"HubRouter is a sub-quadratic routing primitive using learned hubs that replaces attention layers in hybrid models while delivering competitive perplexity and large throughput gains.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"Mamba-2 [8] and Mamba-3 [9] improve SSM expressiveness but acknowledge in- context retrieval limitations [26]. HubRouter provides a learned, content-based alternative for the attention layers in hybrid architectures; we demonstrate it in one hybrid (Jamba-style) and one transformer here, with broader generality to be established in follow-up work. Perceiver and cross-attention bottlenecks.Perceiver [17] uses learned latent tokens that cross-attend to inputs, reducing quadratic cost. HubRouter's encode step is structurally similar but (a) includes decode, score, and council stages for richer processing and (b) supports chunked causal encoding for autoregressive use. Set Transformers [18] use inducing points forO(nM) attention; HubRouter extends this with routing-"},{"citing_arxiv_id":"2604.11095","ref_index":5,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Bottleneck Tokens for Unified Multimodal Retrieval","primary_cat":"cs.LG","submitted_at":"2026-04-13T07:12:12+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Bottleneck Tokens paired with a masked generative objective achieve state-of-the-art unified multimodal retrieval performance among 2B-scale models on the MMEB-V2 benchmark with 78 datasets.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"production, rather than as a structural constraint on how information is com- pressed. The generative loss operates alongside or after embedding extraction, without being architecturally coupled with a dedicated compression mechanism. 2.3 Learnable Tokens for Representation Compression Input-side compression.Learnable tokens are widely used asinput-sidecompres- sors (e.g., Perceiver [5], Q-Former [13], Flamingo [1]) to adapt visual features for LLMs. These modules facilitate modality alignment but do not produce retrieval embeddings. Learnable tokens for embedding extraction.Conversely,output-sidelearnable to- ken approaches are rarer in unified retrieval. NV-Embed [11] introduces a Latent Attention Layer as an explicit replacement for<EOS>pooling."},{"citing_arxiv_id":"2604.09402","ref_index":36,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Enhancing event reconstruction for $\\gamma$-ray particle detector arrays using transformers","primary_cat":"astro-ph.IM","submitted_at":"2026-04-10T15:15:26+00:00","verdict":null,"verdict_confidence":null,"novelty_score":null,"formal_verification":null,"one_line_summary":null,"context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"equipped with two PMTs: one is positioned at the bottom of the upper layer looking upwards, to ensure accurate timing, and the second one is located at the ceiling of the lower layer and is looking downwards (see Figure 1 (b)). The interaction of primary gamma rays and protons with the atmosphere and the result- ing air shower were simulated using CORSIKA [36]. At low energies hadronic interactions are modeled using UrQMD [37], while QGSJet-II.04 [38] is used elsewhere. The interaction of the secondary shower particles with the detector, i.e., the detector response, was simulated with the software package HAWCsim [39]. AERIE [40] andpyswgo2 have been used to perform the 1We will use the terms inner array for the innermost zone and outer array for the remaining zones."},{"citing_arxiv_id":"2602.11229","ref_index":17,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Latent Generative Solvers for Generalizable Long-Term Physics Simulation","primary_cat":"cs.AI","submitted_at":"2026-02-11T15:34:52+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"LGS pretrained on 2.5M trajectories across 16 systems matches deterministic baselines at one step and halves 20-step error while using far less compute and adapting to held-out higher-resolution flows.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2506.14135","ref_index":24,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"GAF: Gaussian Action Field as a 4D Representation for Dynamic World Modeling in Robotic Manipulation","primary_cat":"cs.RO","submitted_at":"2025-06-17T02:55:20+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"GAF creates 4D dynamic scene models by adding motion to 3D Gaussians, enabling better reconstruction and 7.3% higher success in robotic tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2504.06176","ref_index":14,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"A Self-Supervised Framework for Space Object Behaviour Characterisation","primary_cat":"cs.LG","submitted_at":"2025-04-08T16:19:19+00:00","verdict":null,"verdict_confidence":null,"novelty_score":null,"formal_verification":null,"one_line_summary":null,"context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}