arXiv:2606. 05115v1 Announce Type: cross Abstract: Children learn the meanings of words from a continuous, temporally structured stream of egocentric experience.
arXiv:2610.01192v1 Announce Type: new Abstract: Streaming vision-language models must process continuously growing video streams under a bounded compute budget, creating a persistent tension between...
The paper investigates using raw, unlabeled video as a mid‑training phase for a pretrained language model. By encoding video frames into visual tokens and training the model to predict the next token, the authors mid‑train Qwen3‑1.7B on YT‑Temporal‑1B clips. The resulting model outperforms a non‑mid‑trained counterpart by 2.9 points on video benchmarks and 5.1 points on image benchmarks, while maintaining comparable text performance.
VIVAS is a new Vision‑Language Model pre‑training framework that addresses the lack of fine‑grained visual perception in existing VLMs. It introduces a unified token space and a dense‑structural‑semantic vision tokenizer that expands the textual vocabulary with visual tokens, enabling vision‑language unified autoregressive supervision over both visual details and linguistic content. Trained on 12.4 T tokens, VIVAS achieves state‑of‑the‑art results on 7 tasks and 39 multimodal benchmarks.
OneStreamer is a streaming video model that jointly learns to record evidence and respond to tasks through a shared proactive generation process. Its Proactive Hierarchical Caption Memory creates time‑grounded local‑detail captions and event summaries, while Proactive State Transition Learning reduces waiting states by supervising all output anchors. The authors also built a large OneStreamer‑1M dataset and show that a 4B model outperforms baselines on eight streaming video benchmarks, with ablations confirming the benefits of generated captions and PSTL.
arXiv:2509.22650v3 Announce Type: replace Abstract: Most existing approaches to referring segmentation achieve strong performance only through fine-tuning or by composing multiple pre-trained models,...
arXiv:2608.27881v1 Announce Type: new Abstract: Recently, many streaming video understanding methods have been proposed by constructing an external memory to store historical data for computational r...
arXiv:2605. 23045v2 Announce Type: replace-cross Abstract: Video representation learning has seen tremendous progress in recent years.
The paper introduces DynaTokens, a transformer-based token generator that produces fine‑tuning tokens on demand for continual VideoQA with multimodal large language models. It uses shared generation weights and meta‑learning‑inspired regularisers to reduce task interference and forgetting, connecting the objective to sharpness‑aware optimisation for flatter cross‑task minima. Experiments on standard continual VideoQA benchmarks show that DynaTokens achieves higher average accuracy, lower forgetting, better zero‑shot generalisation, and robust cross‑modal transfer in a new ImageQA→VideoQA protocol.
The paper introduces VT-Contrast, a representation-level temporal counterfactual objective designed to improve temporal understanding in Video Language Models (VideoLMs). By supervising late-layer last-frame video tokens and contrasting order-preserving views with reordered counterfactuals graded by Kendall tau distance, VT-Contrast addresses the mismatch between ordered video input and text-based supervision. The method requires no architectural changes, is compatible with various VideoLM training tasks, and demonstrates improved performance on temporal understanding benchmarks.
EgoMemReason is a new benchmark for week‑long egocentric video understanding that focuses on memory‑driven reasoning rather than simple perception tasks. It tests three memory types—entity, event, and behavior—across 500 questions, each requiring evidence from an average of 5.1 video segments and 25.9 hours of backtracking. Evaluation of 17 models shows that even the best achieves only 39.6% accuracy, highlighting the difficulty of long‑horizon memory in multimodal systems.
The paper introduces STITCH, a training‑free method that partitions videos into semantically meaningful temporal chunks using a frozen video‑text backbone. By detecting changes in the embedding sequence of short video windows, STITCH produces reusable temporal abstractions that can be applied to multiple tasks such as event boundary detection, language‑based moment retrieval, and frame selection for vision‑language models. Experiments show that STITCH performs competitively with specialized methods while requiring no task‑specific training, especially when processing is limited to a few frames or tokens.
The paper introduces Temporal Activation Injection (TAI), a post‑processing technique that reinforces temporal representations in Video Large Language Models (VideoLLMs) at inference time. By measuring the temporal divergence vector τ_l— the layer‑wise difference caused by reversing video frame order— the authors find that temporal information peaks at intermediate layers and fades toward the output. TAI injects the peak divergence back into later layers, improving temporal reasoning on three VideoLLMs across four benchmarks without affecting non‑temporal performance.