Smart search ranks by meaning as well as keywords (one row per story, last 45 days).
Audio-visual understanding remains challenging because models must jointly interpret spoken content, visual events, and their temporal relationships. Existing omni models typically introduce dedicated audio encoders and rely on expensive audio-video-text training, tightly coupling omni capability to specific VLM backbones and potentially weakening their existing visual and reasoning abilities.…
Omni-Streaming Thinking fixes premature cross-modal commitment in streaming omni-modal models via pending claims verified against modality-specific evidence, beating baselines by over 10%.
The paper identifies 'premature cross-modal commitment', where streaming models keep relaying early visual interpretations even after audio contradicts them. OST generates evidence-linked pending claims with future verification intervals, stores audio and visual evidence separately, and refutes claims when contradictory evidence appears. Built on a frozen Qwen3-Omni-30B-A3B-Instruct backbone with lightweight adaptation, it outperforms open baselines by more than 10% relative on five streaming and audio-visual benchmarks. On the new OST-DiagBench it reaches d-prime 2.95 versus at most 1.38 for open baselines, while reducing vision-induced auditory hallucinations.
Researchers analyze the 'attention triangle' in audio-video diffusion models, showing bias-driven cross-attention routing causes semantic leakage and proposing inference-time interventions that improve grounding.
A study probes the three cross-attention edges linking text, audio, and video streams in audio-video diffusion models. It finds the audio-video edge is bidirectional and shaped by parameter-encoded biases, so prompts in tension with learned priors can be overridden, producing visually canonical but incorrect outputs. Attention-derived signals are used as diagnostics and to guide inference-time interventions that improve cross-modal semantic grounding while preserving generation quality.
Realtime-Venus introduces two 9B full-duplex interaction models (Omni and Audio) that outperform Gemini 3.1 Live and GPT-4o on continuation metrics.
Realtime-Venus is a proactive full-duplex interaction system built on two separately trained 9B models: Realtime-Venus-Omni for audio-visual interaction and Realtime-Venus-Audio for spoken interaction. A dual-loop runtime lets foreground interaction continue while Realtime-Venus-Harness asynchronously executes background reasoning and tool tasks. Realtime-Venus-Omni leads on six of eight video benchmarks, including StreamingBench (70.2%), OVO-Bench (64.7%), and Daily-Omni (81.3%), while Realtime-Venus-Audio tops MMAU (78.0%) and MMAU-Pro (63.2%). On Full-Duplex-Bench v1.5, Realtime-Venus-Audio handles 75% of interruptions and exceeds Gemini 3.1 Live and GPT-4o on all three continuation metrics.
NVIDIA struck a $12B deal with AI coding startup Poolside, licensing its Model Factory and hiring 109 of its technical employees.
NVIDIA spent roughly $12B in an unusual reverse-execuhire of Poolside, licensing the company's Model Factory while hiring 109 of its ~115 technical staff; founders retain a $1B stake and employees receive about $6B. Poolside had raced to raise $2B to fund a 40,000 GB300 cluster after missing a six-week funding window, and founders argue frontier-scale training now requires an order of magnitude more compute plus contracted data center space. An infrastructure arm spun out in January 2026 is scaling toward 7GW as a neocloud. The newsletter also recaps OpenAI and Anthropic agent-platform releases.
A survey catalogs inference-efficiency techniques for video and audiovisual LLMs, mapping bottlenecks in sampling, encoding, token reduction, and LLM decoding.
This survey covers inference-efficiency mechanisms for visual and audiovisual video LLMs, reporting reductions in parameters, FLOPs, latency, memory, and token counts. It organizes methods by pipeline stage, covering frame sampling, modality encoding, connector-level token reduction, and LLM prefilling and decoding for systems built since late 2022. The authors compile accuracy-cost comparisons under shared host models and input protocols, identify gaps in audiovisual efficiency and standardized evaluation, and maintain a public repository.
Researchers introduce CVRR, forcing multimodal models to rely on recurrent latent computation rather than accessible image tokens, validated via causal interventions and benchmarks.
The paper presents Causal Visual Recurrent Reasoning (CVRR), which makes recurrent hidden-state computation the required image-conditioned path for prediction in vision-language models. Before decoding, visual states and the original multimodal KV cache are removed so only the final recurrent state carries image information to the answer. CVRR retains strong performance on V*, MMVP, BLINK, and MME-RealWorld-Lite while comparable latent reasoners fail under the same constraint. Causal interventions show predictions remain sensitive to recurrent content and that persistent visual evidence causally revises the recurrent trajectory.
Researchers propose Vision-of-Thought (VoT), a discrete visual-planning token layer between VLMs and diffusion transformers improving text-to-image semantic alignment.
VoT introduces a discrete visual-thinking layer between vision-language models and diffusion transformers, letting the VLM act as a multimodal planner that emits tokens describing objects and layouts before pixel generation. A specialized VoT tokenizer is trained with VLM alignment, feature reconstruction, and vector-quantization losses. Experiments show improved semantic alignment and a structured, interpretable interface for controllable generation.
AlayaVista is a camera-controllable streaming video world model that decouples panoramic scene evolution from perspective synthesis, trained on a 1,318-hour 4K dataset.
AlayaVista builds a 360-degree scene prior from a single perspective image, evolves it as a camera-conditioned panoramic latent state, and maps it to perspective video via a latent viewport renderer plus a perspective refiner. Chunk-autoregressive generation and few-step distillation enable efficient streaming. The authors introduce MUGEN, a real-world panoramic video dataset with 1,318 hours of at-least-4K video and rich semantic and geometric annotations.
A controlled pure-autoregressive testbed shows task-specific validation losses rank image tokenizers differently, with I2T loss the most consistent signal.
Researchers built a controlled pure-autoregressive testbed and tracked task-specific validation losses during multimodal continual pretraining across text, image, text-to-image (T2I), and image-to-text (I2T) prediction. They find losses should be analyzed per task because they exhibit distinct scaling behavior and rank tokenizers differently, and that the loss-performance relationship depends on the predicted token space. I2T loss, computed over a shared text vocabulary, correlates consistently with both generation and visual understanding performance after supervised finetuning. Case studies revisit the discriminator, semantic supervision, and vocabulary size as tokenizer design axes.
LynnReal-Omni unifies controllable video generation tasks in a 32B multimodal diffusion transformer, with a 27B Flash variant rendering 540p clips in 377 ms.
LynnReal-Omni is a native multimodal video generation framework built on a 32B shared multimodal diffusion transformer unifying text-to-video, image-conditioned generation, reference guidance, structural control, editing, restoration and long-video generation, accepting heterogeneous inputs like 3D renders and game recordings for agentic visual workflows. A dedicated 27B Flash model enables real-time rendering, producing a 22-frame 540p video in 377 ms on one H100 versus 843 ms for the full model. The work introduces a curated multi-shot audiovisual data pipeline and MSAVP, a 100-prompt, 20-metric evaluation design covering instruction following, plausibility, visual quality, temporal behavior and audio coordination.
SenseTime releases SenseNova-U1.5, an 8B-MoT encoder-free multimodal model unifying visual understanding, reasoning, and generation with native 4K resolution.
SenseNova-U1.5 is an 8B mixture-of-transformers multimodal model with an encoder-free, VAE-free architecture that understands, reasons about, and generates visual content at native resolutions up to 4K. Post-training optimizes specialized experts for visual aesthetics, bilingual text rendering, infographic generation, and image editing, consolidated through multi-expert on-policy distillation. Evaluations report gains in image fidelity, text rendering, multi-reference editing, and instruction following. The team plans to open-source training code including supervised fine-tuning, reinforcement learning, and on-policy distillation.
Researchers diagnose conditional visual grounding failures in visuomotor imitation policies and show targeted interventions substantially improve distractor robustness.
The paper studies why ACT-based visuomotor imitation policies fail when visually similar distractor objects or receptacles are introduced, finding sensitivity depends on both distractor type and manipulation stage. Interventions including distractor augmentation, phase-dependent attention regularization, and appearance-based visual prompting improve target selection while preserving spatial control information, with gains in simulation and on a physical UR3e. The same failure pattern is confirmed in a pretrained vision-language-action policy on a state-conditioned medical instrument-handling task.
StepAudio 3 Gen unifies TTS, voice design, music, and sound effects via discrete autoregressive modeling over RVQ tokens.
StepAudio 3 Gen is a general-purpose audio generation model covering zero-shot TTS, voice design, vocal generation, sound effects, music, vibe speech, and mixed audio in one framework. It uses discrete autoregressive modeling over residual vector quantization (RVQ) tokens rather than the diffusion Transformer paradigm, with a StepAudio Tokenizer representing audio at 12.5 Hz in a shared 16x2048 residual code space. Key design principles include interference-aware progressive pretraining, an RVQ Adaptor for multi-codebook acoustic representations, and shared discrete autoregressive modeling. The model reports state-of-the-art performance on TTS and voice design while retaining strong generation across speech, vocals, sound effects, and music.
SimpleMemVLA passes full timestamped video history straight to a VLA backbone, setting state of the art on four memory benchmarks.
SimpleMemVLA is a vision-language-action model for long-horizon manipulation that removes the dedicated memory module entirely. It keeps sampled history intact and feeds it to the backbone as timestamped video, with the hidden states of a generated sub-task serving as the only channel into a standard flow-matching action head. Prefilling the shared history prefix during action execution keeps latency close to a single-frame VLA. The system sets a new state of the art on four memory benchmarks and outperforms retrieval, compression and recurrent-state mechanisms, with causal interventions confirming the policy genuinely reads its history.
A controlled autoregressive testbed shows validation losses must be analyzed per task, and image tokenizer choice affects joint multimodal text modeling.
Researchers built a pure-autoregressive testbed to study image tokenizers as the 'visual language' of unified multimodal models, tracking task-specific validation losses during multimodal continual pretraining across text, image, text-to-image (T2I), and image-to-text (I2T) prediction. They found that losses exhibit distinct scaling behavior per task and rank tokenizers differently, and that I2T loss over a shared text vocabulary gives a more consistent loss–performance signal than T2I loss. Better reconstruction does not necessarily yield lower task-specific losses or stronger downstream performance, and tokenizer choice can affect text modeling under joint optimization. Case studies examine the discriminator, semantic supervision, and vocabulary size design axes.
RenderFormer-V2 is a transformer-based neural renderer handling caustics, volumetric scattering and out-of-distribution materials without per-scene training or specialized code.
RenderFormer-V2 is a learned transformer-based neural rendering model that models global light transport as a sequence-to-sequence transformation, handling caustics, volumetric scattering, environment lighting, textured and displaced surfaces and out-of-distribution materials. It uses a two-stage process: a view-independent stage resolving primitive-to-primitive transport, and a view-dependent stage converting the neural scene representation into pixels. Improvements include combined windowed-attention with a rendering-informed attention sink for scalability, support for heterogeneous primitives like environment maps and participating media, and a surface-reflectance-independent neural material encoding, validated across diverse scenes with extensive ablations.
Researchers present Diffusion TV, a CRT-based installation where antenna manipulation lets audiences physically experience diffusion model denoising.
Diffusion TV is an interactive installation built around a modified CRT television where turning the antenna controls the clarity of AI-generated images and sounds, mirroring the denoising process of diffusion models. Three channels present AI-generated animals from the past, present, and future within a temporal and ecological narrative. The authors frame the work as an embodied, non-verbal alternative to explainable AI that highlights intermediate generative states rather than final outputs.
A developer released a browser-based tool that visualizes which past tokens influence each LLM output token using aggregated, value-weighted attention scores.
A Show HN project presents a React application built on Transformers.js that renders per-token attention influence by aggregating attention weights scaled by value-vector magnitudes across all attention heads and layers. To expose internal tensors, the author instrumented the ONNX computation graph, hosted a modified model on Hugging Face, and pre-generated prompts to avoid long model downloads in the browser. Demos with a 600-million-parameter model show how verbatim copying draws heavily on source tokens and how single outputs blend information from multiple phrases.
Ambient team wins EgoLongQA 2026 sub-2B division by distilling an agentic long-video perception pipeline into a 2B vision-language model.
Ambient's entry to the EgoLongQA track of the Wearable-AI Challenge at ECCV 2026 placed first in the <=2B parameter division with 0.8279 on the held-out test set. The system distills the junior perception module of a tool-using agentic pipeline into a 2B student, reaching 89% of the pipeline's accuracy with 1.1% of its parameters and lifting a 27.1% base model to 81.4%. To meet the division limit, the multilingual embedding table is pruned from 248,320 to 143,469 rows, reaching 1.9985B parameters with provably identical logits on retained rows.
Vidu S2 delivers real-time 720p interactive avatar generation plus real-time video editing with style, clothing, character, and background replacement.
Vidu S2 comprises Vidu S2-Avatar, a real-time interactive digital-character model, and Vidu S2-Editing, a real-time video editing model. S2-Avatar supports real-time 720p generation, dynamic references updatable at any moment, and stronger instruction following such as dancing, while S2-Editing performs real-time style rendering, clothing replacement, character replacement, and background replacement. The work also explores real-time spatial video generation for both models, reports outperforming all baselines, and offers a playable online demo at vidu.com.
Ref-GeNVS is a training-free method for reflection-consistent generative novel view synthesis that treats mirror images as two complementary views.
An arXiv paper proposes Ref-GeNVS, a training-free approach to generative novel view synthesis in scenes containing mirrors. It estimates the mirror plane, reflects camera poses to create virtual views, and applies mirror-gated attention plus reflection injection within a multi-view diffusion model. On synthetic and real mirror scenes, Ref-GeNVS outperforms recent generative NVS methods while requiring no fine-tuning.
Blender-VideoBench evaluates agentic video understanding by having agents programmatically reconstruct real videos in Blender scenes.
BVB (Blender-VideoBench) tests whether multimodal agents truly understand videos by requiring programmatic reconstruction of real-world videos as animated Blender scenes via a lightweight Mini-BVB harness under identical sandbox and cost constraints. Evaluation uses Dual VQA for spatiotemporal fact preservation and Latent Similarity for perceptual match, combined in a square-root mean overall score. Across 51 configurations from 10 model families, the best model reaches 88.6 Latent Similarity but retains only 53.7% of source-correct spatiotemporal answers, showing semantic retention remains the main challenge.
Amazon Prime Video launched AI lip-sync dubbing aligning actors' mouth movements with human-dubbed audio, starting with the English dub of Maxton Hall.
Prime Video's new feature combines AI and visual effects technologies to make lip movements match translated speech, currently available for seasons 1-2 of the German series Maxton Hall worldwide and planned for additional titles, with season 3 arriving December 9. Amazon previously piloted AI-aided dubbing in English and Spanish across 12 movies and shows last year. Meta and YouTube have recently launched similar AI auto-dubbing features with optional lip sync for creators.
StepAudio 3 Realtime debuts an audio-language model with Think-While-Speaking reasoning, delivering full-duplex voice dialogue with top benchmark results.
StepAudio 3 Realtime is an audio-language foundation model built around a continuous listen-converse-think-act loop for real-time spoken interaction. Think-While-Speaking runs private reasoning in parallel with speech, reaching a 73.0 macro average on StepAudioChat in reasoning mode. The model reports 90.6 on MMSU, 98.9 overall on the Artificial Analysis Full-Duplex Bench, and 56.0% macro task success on tau-Voice. An integrated Voice Agent handles asynchronous tool execution without disrupting dialogue flow.
Viggle released Viggle-Animate, a 33.1B MiniMax-H3 finetune replacing video characters from one repainted frame, rendering 124 frames in 26 seconds on one GPU.
Viggle-Animate replaces the character in a video using only a driving video and one of its own repainted frames, with no pose estimator, segmentation mask, face tracker, or text encoder. It is a 33.1B full finetune of MiniMax-H3's ref2va transformer, jointly distilled with DMD across two teachers split by noise level, so rendering takes three forward passes per clip. On a B200 GPU it renders 124 frames in 26 seconds, 6.1x faster per clip than Wan2.2-Animate-14B in matched comparisons. The method assumes no person-specific representation, so it generalizes beyond humans; a demo, research write-up, and ComfyUI nodes are available.
Researchers introduce KoNA, a benchmark exposing vision-language models' failures at selective non-compliance, plus fine-tuning that improves refusal and abstention accuracy.
KoNA is a benchmark for evaluating selective non-compliance in vision-language models across five categories: False Premise, Visual Inaccessibility, Universal Unknown, Task Feasibility and Safety. It tests both query-level and component-level non-compliance using paired single and compound queries, and evaluations across diverse VLMs show models often fail to refuse, correct or abstain appropriately, with failures worsening on compound queries. Fine-tuning VLMs on KoNA examples substantially improves non-compliance accuracy while largely maintaining performance on fully answerable tasks.
ECHO benchmark pairs identical-overlap Chinese dialogue examples with contrasting contexts, revealing most full-duplex systems bias toward yielding the floor.
ECHO is a paired diagnostic benchmark for Chinese full-duplex turn-taking that matches examples with identical overlap transcripts but contrasting preceding multi-turn contexts, one requiring Yield and the other Keep. It also includes off-talk examples for diagnosing unnecessary yielding and introduces pair accuracy, which grants no credit to constant-action policies. Experiments across multiple full-duplex systems show most exhibit a pronounced Yield bias, performing substantially better on interruptions than backchannels. The benchmark and metadata will be publicly released.
Probing study shows vision encoders make canonical color linearly decodable from grayscale images and tie it to object identity.
Researchers use canonical color as a controlled testbed for measuring conceptual (not just visible) information in vision encoder representations. A dataset of objects with canonical colors was built, and probes on both color and grayscale images show canonical color remains decodable even when color is removed from the input, linked to predicted object identity. Extending to full VLMs, they find post-training has a surprisingly large effect on color decodability in the vision encoder.
CARDEA, a vision-language model trained only on public data, matches cardiologists on coronary angiography complexity assessment while exposing auditable bounding-box evidence.
CARDEA is a unified large vision-language model serving as the inference core of an end-to-end coronary angiography pipeline from multi-view videos to study-level diagnosis. It was trained on public datasets through visual alignment, self-distilled Chain-of-Box cold start, and reinforcement learning with verifiable rewards encouraging bounding-box reasoning. It reached 0.91 accuracy on dominance classification under domain shift and 0.90 on complexity assessment, comparable to two interventional cardiologists. RLVR raised zero-shot report generation vessel-severity macro-F1 from 0.513 to 0.686, while supervised imitation alone did not.
Apple Watch Series 12 and Ultra 4 add opt-in audio intelligence features that process microphone audio on-device via a new Secure Exclave.
The Apple Watch Series 12 and Ultra 4 ship with four opt-in audio intelligence features: Sound Recognition, Shazam music identification, Siri Recap conversation summaries, and Live Rewind 15-second transcription. Audio is held and processed in an isolated Secure Exclave buffer on the new S11 chips, with on-device speech recognition on iPhone producing a distilled transcript that foundation models in Private Cloud Compute then summarize. Apple says no raw audio is stored or accessible to the operating system, apps, the user, or Apple, and untransferred audio is automatically deleted.
VDiff-Bench, a 1,756-question benchmark, shows multimodal LLMs struggle with fine-grained image-difference identification, scoring as low as 8.7% on low-level changes.
VDiff-Bench is a multiple-choice benchmark of 1,756 four-way questions over image pairs covering 10 change categories including position, motion, color, texture, OCR/text and illumination, with curated hard negatives. Evaluation of 11 state-of-the-art open- and closed-source MLLMs shows fine-grained visual comparison remains brittle: 7-8B-scale open-source models score 52.5-70.6% on semantic changes but only 8.7-33.3% on low-level changes like noise and texture. Notably, Grok 4.3 shows a sharp performance drop on noise and texture differences, falling behind large open-source models like Kimi K2.5 and K3.