braintankAIhttps://braintank.ai/Frontier technology, read from the paper.en Robot control model anticipates motion without generating future videohttps://braintank.ai/a/arxiv-2610.37381.htmlhttps://braintank.ai/a/arxiv-2610.37381.htmlSun, 01 Nov 2026 09:00:00 +0900EmbodiedA preprint describes a vision-language-action model that learns compact physical tokens about how scenes will change, sidestepping the delay of generating video. Tencent's Hy-Nomas generates video worlds that respond to text and camera commandshttps://braintank.ai/a/arxiv-2610.36729.htmlhttps://braintank.ai/a/arxiv-2610.36729.htmlSun, 01 Nov 2026 09:00:00 +0900EmbodiedA video world model from Tencent Hunyuan follows successive typed instructions and camera moves while keeping track of what has already happened on screen. Surface-shape maps help an AI video model keep a subject consistent from every sidehttps://braintank.ai/a/arxiv-2610.35684.htmlhttps://braintank.ai/a/arxiv-2610.35684.htmlSun, 01 Nov 2026 09:00:00 +0900LatentA KAIST AI preprint steers video generation with geometry maps rebuilt from reference photos, so a subject stays the same as the camera circles it. Leaner streaming model picks one voice out of crowded, noisy sceneshttps://braintank.ai/a/arxiv-2610.32749.htmlhttps://braintank.ai/a/arxiv-2610.32749.htmlSun, 01 Nov 2026 09:00:00 +0900LatentA preprint offers a more realistic test for audio-visual voice extraction, plus a streaming model the authors say is far smaller and faster than an existing baseline. Teaching diffusion vision-language models to look at the image only when neededhttps://braintank.ai/a/arxiv-2610.36671.htmlhttps://braintank.ai/a/arxiv-2610.36671.htmlSun, 01 Nov 2026 09:00:00 +0900LatentA decoding method from KAIST AI checks whether the picture would change the answer before processing it, cutting reported decoding computation to about a third. Single-cell map links rare and common DNA variants to immune gene activityhttps://braintank.ai/a/W7224109065.htmlhttps://braintank.ai/a/W7224109065.htmlSun, 01 Nov 2026 09:00:00 +0900MorphogenSequencing more than five million immune cells from 1,925 people shows how inherited variation tunes gene expression, often in just one or a few cell types. Brain scans of 2,175 people find no single neural signature of emotional controlhttps://braintank.ai/a/W7224107502.htmlhttps://braintank.ai/a/W7224107502.htmlSun, 01 Nov 2026 09:00:00 +0900InteriorA preregistered analysis of 40 brain-imaging datasets finds that common measures of how well people reframe their feelings disagree, and none tracks prefrontal activity. Inherited DNA variants change how open the genome is, cell type by cell typehttps://braintank.ai/a/W7224084810.htmlhttps://braintank.ai/a/W7224084810.htmlSun, 01 Nov 2026 09:00:00 +0900MorphogenA study ties genetic variation to single-cell chromatin accessibility and finds that the genetic signal helps an AI model map how genes are switched on. Human insula divides into four regions, with the front one steering the othershttps://braintank.ai/a/W7154962243.htmlhttps://braintank.ai/a/W7154962243.htmlSun, 01 Nov 2026 09:00:00 +0900InteriorStimulation and recordings in 87 neurosurgical participants map four functional regions of the insula and an asymmetric pathway through which cognition may shape pain. Dedicated memory for AI agents often underperforms simple retrieval, study reportshttps://braintank.ai/a/arxiv-2610.34073.htmlhttps://braintank.ai/a/arxiv-2610.34073.htmlSun, 01 Nov 2026 09:00:00 +0900AgencyA large-scale evaluation suggests that the way memory systems reorganise an agent's history can hide evidence, a cost the authors call memory tax. Benchmark finds AI agents remember conversations but struggle to infer user habitshttps://braintank.ai/a/arxiv-2610.24520.htmlhttps://braintank.ai/a/arxiv-2610.24520.htmlSun, 01 Nov 2026 09:00:00 +0900AgencyA test built on hundreds of synthetic sessions finds the best agent memory method answering only about a third of habit questions correctly. AI modelled on a panel of doctors beats generalist physicians in a diagnosis benchmarkhttps://braintank.ai/a/W4417175683.htmlhttps://braintank.ai/a/W4417175683.htmlSun, 01 Nov 2026 09:00:00 +0900LatentIn a benchmark where diagnosticians must decide what information to gather, a language-model system far outscored generalist doctors, under controlled conditions. Atomic defects pin electrons in place inside a two-dimensional metalhttps://braintank.ai/a/W4404313154.htmlhttps://braintank.ai/a/W4404313154.htmlSun, 01 Nov 2026 09:00:00 +0900SubstrateMicroscopy reveals two localised electronic states hiding among a metal's mobile electrons, an electronic counterpart to an effect already achieved with light. X-rays reveal handed spin waves in manganese telluride, a candidate altermagnethttps://braintank.ai/a/W4406975366.htmlhttps://braintank.ai/a/W4406975366.htmlSun, 01 Nov 2026 09:00:00 +0900SubstratePolarised X-ray scattering uncovers a magnetic ripple with a definite twist, whose spin polarisation varies with momentum in line with the material's symmetry. Robots learn faster by rewinding imagined futures inside a world modelhttps://braintank.ai/a/arxiv-2610.36068.htmlhttps://braintank.ai/a/arxiv-2610.36068.htmlSun, 01 Nov 2026 09:00:00 +0900EmbodiedA preprint describes a training method that replays a robot's decisions in a learned simulator, tries alternatives, and credits the choices that made the difference. AI that judges left from right may not learn how far away things arehttps://braintank.ai/a/arxiv-2610.34283.htmlhttps://braintank.ai/a/arxiv-2610.34283.htmlSun, 01 Nov 2026 09:00:00 +0900LatentA preprint names a training pitfall called interface lock-in and reports that replaying mixed spatial data in time order improves video models on three benchmarks at once. Tencent's single model parses documents, images, audio and video into one formathttps://braintank.ai/a/arxiv-2610.39301.htmlhttps://braintank.ai/a/arxiv-2610.39301.htmlSun, 01 Nov 2026 09:00:00 +0900LatentA preprint describes one shared encoder and one output schema for ten parsing tasks, with the authors reporting leading scores on document and multimodal benchmarks. A scheduler that plans around AI agents' tool calls cuts serving delayshttps://braintank.ai/a/arxiv-2610.37230.htmlhttps://braintank.ai/a/arxiv-2610.37230.htmlSun, 01 Nov 2026 09:00:00 +0900LatentTAO, described in a preprint, picks which agent sessions get GPU time by weighing when their tools will return, what is cached where and how many tokens they carry. Self-training AI decides for itself when its practice tasks have gone stalehttps://braintank.ai/a/arxiv-2610.38568.htmlhttps://braintank.ai/a/arxiv-2610.38568.htmlSun, 01 Nov 2026 09:00:00 +0900LatentA preprint from Xiaomi Research describes a training loop that watches its own learning signal and swaps in fresh, checkable questions, with no human-labelled data. Imagined futures let robot models learn from human videos without action labelshttps://braintank.ai/a/arxiv-2610.37666.htmlhttps://braintank.ai/a/arxiv-2610.37666.htmlSun, 01 Nov 2026 09:00:00 +0900EmbodiedThinkWorld, a preprint framework from NVIDIA-affiliated researchers, rewards a robot model when the imagined outcome of a planned motion matches what real footage shows.