Technology & AI
AI vision models encode correct count but output wrong numbers
Vision-language models internally encode the right answer but misread it, a new arXiv study finds. A simple fix lifts counting accuracy 15.6 points.
AI Models
AI model releases and benchmarks — what each new model actually changes for the people who use it.
191 stories
Technology & AI
Spotify study links narrator voice to audiobook appeal
Spotify and Queen Mary researchers found voice features like tone and pace correlate with audiobook engagement even after controlling for the title.
Technology & AI
Dynamic Fréchet Regression predicts distributions, selects features
A new arXiv preprint introduces Dynamic Fréchet Regression, predicting full probability distributions across an index and selecting driving features.
Technology & AI
MedRealMM benchmark: AI matches doctors but fails on safety
New 5,620-case benchmark finds frontier AI matches physicians on positive clinical steps but triggers more unsafe responses in real consultations.
Technology & AI
Knowledge graphs and XAI reshape urban mining audits
New DFKI preprint argues AI in pre-demolition audits must prioritise defensible decisions over accuracy, proposing four knowledge-graph modes.
Technology & AI
Physics-constrained ML cuts combustion simulation cost 10x
Physics-constrained ML encodes the second law of thermodynamics into a combustion surrogate, cutting simulation costs by more than 10x.
Technology & AI
LLM agent coordination framework cuts communication 70x
LDT-Coord uses a lightweight digital twin to coordinate teams of LLM-powered robots, matching conventional methods at a fraction of the chatter.
Technology & AI
Malaika AI framework uses tri-grounded reasoning for malware analysis
Malaika, a multi-agent AI framework posted to arXiv this month, uses three grounding mechanisms to make malware analysis more precise and auditable.
Technology & AI
TrustX ARC rates AI agent risk across 12 dimensions
TrustX ARC, a new arXiv preprint, scores agentic AI risk across 12 dimensions into three governance tiers for risk officers, developers and regulators.
Technology & AI
LLM agent framework blocks hallucinated actions in industrial control
A new arXiv preprint pairs an LLM planner with a forecasting model to guard industrial control systems, recording zero hallucinated actions in attack
Technology & AI
KV-PRM cuts AI agent scoring cost 5,000x via cache reuse
KV-PRM reads the memory AI models already produce during generation, slashing the cost of verifying multi-agent reasoning chains by up to 5,000x.
Technology & AI
CogniConsole: LLM reliability tied to control, not capability
New arXiv paper with 489 probes shows structural scaffolding around a fixed LLM cuts failure rates, challenging the bigger-model assumption.