Think in 360°
PanoVideoReasoner reaches 55.7% on PanoVideoBench, improving its backbone by 12.5 points and the previous SOTA by 2.6.
Benchmarking and Bootstrapping Panoramic Video Reasoning
October 31 – November 6: 18 frontier papers out of 2,440 read.
Papers rated as moving the field
How we choosePanoVideoReasoner reaches 55.7% on PanoVideoBench, improving its backbone by 12.5 points and the previous SOTA by 2.6.
Benchmarking and Bootstrapping Panoramic Video Reasoning
On FrontierPhysics, no AI configuration among 26 passes more than 46 of 168 end-to-end physics research rollouts.
Structured-output backdoors inserted via Kerkoporta reach success rates of up to 96.2% in reasoning coding agents.
Movens-Zero shows consistent gains across text, speech, music, and video when scaling execution-verified data and action-expert capacity.
On SAC-Bench, Event Identity accuracy is 24.8--46.5 percentage points lower under conflicting claims than compatible ones.
DexArena introduces 10 long-horizon and 30 stage-level bimanual dexterous manipulation tasks, with baselines revealing substantial challenges in precision, generalization, and skill composition.
Asking for citations raises fabrication sharply in post-trained checkpoints but not base ones, a 47.8-point interaction.
On the Multi-step SWE benchmark, the strongest of nine frontier models solves only 18.5% of 69 tasks.
A behavioural framework's loss-of-control score correlates with failure rates in contained long-horizon risk scenarios across 13 frontier models (Pearson r = 0.84).
“We measure this at the generation stage across six models and ten consumer-service domains, for three attack families, and show that it transfers to production browser agents, outperforming instruction injection, and to a retrieval pipeline (1.0-57.0% end-to-end success, depending on the retriever).”
BrainwashedLLM achieves ASR@5 between 0.74 and 0.98 across four state-of-the-art LRMs without querying targets during attack construction.
Safin-1.5, trained through recursive safety improvement with a formal reasoning harness, outperforms frontier models in safety and reasoning.
“Results show that current models can follow coarse experimental procedures in videos, but still struggle to ground actions into operable regions and feasible trajectories, to recognize fine-grained laboratory apparatus, states, and safety norms, and to order steps in mid-length procedures.”
PrimeSim leads four of five WorldSimProbe suites and posts the best overall EWMBench score against five open-source action-conditioned world models.
Reward-based alignment training can only encode norms as prices charged when violations are noticed, favoring conditional compliance.
AgiWorld builds persistent worlds from a single image and free-form instruction, achieving state-of-the-art performance on world revisiting and exploration benchmarks.
Leaked graders in agentic environments show GPT models score-seek, violating instructions at similar rates regardless of score use.
Oct 31 – Nov 6: 285 cleared, 18 frontier
Papers this week that name the model
Papers this week with an author at the lab
267 papers, by field