arXiv:2610.34926v1 Announce Type: new Abstract: An action seen once and the same action seen three times receive the same action-recognition label, so that supervision does not require a model to cou...
An action seen once and the same action seen three times receive the same action-recognition label, so that supervision does not require a model to count. Answers shared across histories that differ i...
arXiv:2609.39195v1 Announce Type: new Abstract: Spatial intelligence requires maintaining a coherent understanding of the world as the embodied agent moves. Like humans, the agent must use its own mo...
OVO‑S‑Bench is a fully human‑annotated benchmark designed to evaluate streaming spatial intelligence in multimodal large language models (MLLMs). It contains 1,680 questions derived from 348 source videos, each with a query timestamp and evidence interval, and tests models on four levels of abstraction: instantaneous egocentric perception, spatiotemporal context tracking, generative spatial reasoning, and allocentric spatial mapping. Across 38 MLLMs, Gemini‑3.1‑Pro scored 59.2 versus 92.2 for human experts, with allocentric spatial mapping identified as the main challenge, and the benchmark reveals that chain‑of‑thought reasoning can worsen spatial errors when not grounded in the stream.
Anvil is a unified framework for embodied spatial intelligence that integrates data generation, training, and benchmarking. It introduces three components: Anvil-Gen, a heterogeneous data flywheel that generates and verifies cross‑stage consistency; Anvil-Ben, which provides scene‑aligned supervision and assessments to separate stage competence from error propagation; and Anvil‑MOPD, a 9B student model that distills spatial knowledge from specialist teachers using semantic‑first fault routing and stage normalization. The system builds 5.8 M source‑grounded records, trains on 5.22 M, and evaluates on 580 K, outperforming state‑of‑the‑art models on seven public benchmarks.
arXiv:2609.00291v1 Announce Type: new Abstract: Streaming video understanding requires answering questions that arrive at arbitrary moments over an unbounded video stream. Existing systems primarily...
The paper investigates why constrained updates in supervised fine‑tuning (SFT) still lead to forgetting by developing a functional‑path analysis that decomposes changes in old loss into controlled direction, gradient staleness, probe mismatch, and finite‑step effects. It introduces a conditional refresh budget and identifies a restoring direction provided by informative replay, showing that the placement of constraints relative to Adam and the granularity of constraints significantly affect retention across code‑label audits. The study also explores checkpointed generation, replay schedules, and sequential‑task comparisons, demonstrating that subspace regularization, projection, and supervised replay differentially influence backward transfer and content retention.
arXiv:2610.15407v1 Announce Type: new Abstract: An end-to-end success rate alone cannot distinguish failure to learn an action from failure to apply it as task state changes. We study this distinctio...
arXiv:2610.16632v1 Announce Type: cross Abstract: Memory-based long-video generation aims to combine user control with long-range visual consistency. We organize this process into three functional co...
The paper introduces RePlanMem-VLA, a selective visual memory system for Vision‑Language‑Action models that uses a policy’s own plan revisions as an online memory‑write signal. By detecting unusually large disagreements between temporally aligned action predictions, the method identifies task‑relevant events and updates a bounded visual memory while suppressing redundant frames. Experiments on MIKASA‑Robo‑VLA tasks and real‑robot replay data show that RePlanMem‑VLA retains essential observations in every episode and achieves high ROC‑AUC scores for event detection.
The paper investigates whether progress on spatial reasoning benchmarks translates into improved navigation performance. It finds a gap between benchmark-oriented spatial specialization and real navigation, and proposes aligning spatial supervision with navigation goals, phases, and decision learning. The authors introduce Spatial-Nav-100K, a two-stage fine-tuning approach, and Spatial-NPD, a teacher–student framework that eliminates explicit spatial reasoning at inference, achieving state‑of‑the‑art success rates on HM3D and MP3D with efficient GPU usage.
arXiv:2609.40055v1 Announce Type: cross Abstract: On-policy distillation (OPD) provides dense supervision directly on student-generated trajectories, making it an effective post-training strategy for...