A new AI-powered framework called InqEduAgent uses a Gaussian process to analyze students' cognitive and evaluative traits, then pairs them with compatible study partners to improve collaborative lear…
A new AI-driven dual-attention deep reinforcement learning system reduces scheduling delays in prefabricated construction by reaching within a 4% margin of constraint-programming references, outperfor…
ColaUntangle, a new framework using large language models (LLMs) to untangle software commits, achieved 44% improvement on 1,612 C# tangled commits and 82% on 14,000 Java tangled commits in tests. The…
A new study proposes an active policy selection method for offline-to-online reinforcement learning that uses upper-confidence bounds to allocate limited online interactions between evaluation and fin…
Search APIs from Brave, Tavily, and Firecrawl serve as decision surfaces for AI agents, not just retrieval tools, according to an experiment using a fixed GPT-5.4 agent and 100 SEALQA-HARD questions. …
Generative social agents (GSAs) are poised to transform human-AI interaction through the new Knowledge-based Persuasion Model (KPM), which enables ethical persuasion by leveraging knowledge of self, u…
Researchers have experimentally validated a new attack technique called Trojan Horse Prompting that exploits AI models' trust in their own conversational history, achieving a higher attack success rat…
A new explainable agentic system for detecting conversational scams achieved 97.8% accuracy on the ConScamBench-278 benchmark and identified all 83 scams in the LoveFraud02 corpus, according to resear…
The Department of Housing and Urban Development (HUD) has withheld documents related to DOGE's AI activities, citing a privilege that does not officially exist, raising concerns about AI transparency …
A new AI model called GenGNN challenges the need for complex Graph Transformers in discrete graph generation, achieving up to 5x faster inference times with a simpler message-passing approach. Develop…
A new framework automates quantum circuit design using graph-based Bayesian optimization and a graph neural network surrogate, achieving competitive classification accuracy on cybersecurity data with …
Researchers have developed a transformer-based imitation learning framework for soft robotic guidewire navigation, achieving an 83% success rate on unseen vascular geometries and 75% on a patient-deri…
A new benchmark, VoxENES 2026, reveals that current audio spoofing detectors are largely ineffective against modern synthetic speech, with the best model achieving only a 28.98% Equal Error Rate (EER)…
Researchers have introduced identity bridge supervision to overcome the curse of two-hop reasoning in large language models, enabling out-of-distribution generalization even in simple one-layer transf…
Large Language Models (LLMs) still struggle with social reasoning, according to a new benchmarking task called the Epistemic Asymmetry Schelling Task (EAST). The EAST reveals significant gaps in LLMs'…
Adversarial World Modeling (AWM), a new framework for training autonomous vehicle planners, uses a min-max game and multi-agent self-play to enhance performance in rare, safety-critical traffic scenar…
Diachronic Sample Integration (DSI), a new approach for deep generative models, reduces tail-estimation errors in rare event prediction by aggregating samples across training checkpoints, outperformin…
Researchers have developed a language-guided vision-AI framework for waste management that integrates visual and linguistic reasoning, achieving a BLEU score of 0.8291 and a BERTScore of 0.9273 on the…
AgentCheck, an open-source platform for testing AI tools under real-world failure conditions, lets developers simulate 12 types of faults including timeouts and stale data before deployment. Across fi…
Researchers have introduced commit-time authorization as a new security standard for LLM agents, ensuring durable effects like database updates or sent emails are only committed if authorization condi…