AI

Mitigating Catastrophic Forgetting in Target Language Adaptation of LLMs via Source-Shielded Updates

Aerial Vision-Language Navigation with a Unified Framework for Spatial, Temporal and Embodied Reasoning

Strategic Response of News Publishers to Generative AI

Two Pathways to Truthfulness: On the Intrinsic Encoding of LLM Hallucinations

Visual Self-Fulfilling Alignment: Shaping Safety-Oriented Personas via Threat-Related Images

SelfGrader: Stable Jailbreak Detection for Large Language Models using Token-Level Logits

MixAtlas: Uncertainty-aware Data Mixture Optimization for Multimodal LLM Midtraining

OpenAI’s Internal Memo Attacks Anthropic Over Revenue

Google's Latest AI Update Makes Industrial Robots Way Smarter—Here's How

Allbirds stock moons 800% as it swaps sneakers for AI cloud

Allbirds, the sustainable footwear company, has reportedly pivoted toward AI cloud services, resulting in an 800% stock surge. The strategic shift reflects the company's response to financial difficulties and signals a potential rebranding away from its core sneaker business into the booming AI infrastructure sector.

After sale of its shoe business, Allbirds pivots to AI

Allbirds, the sustainable footwear company, has sold its shoe business and rebranded as NewBird AI, securing $50M in convertible financing to pivot toward AI infrastructure. This dramatic strategic shift reflects broader market pressures on traditional consumer brands and the influx of capital into AI-related ventures.

Hybrid-AIRL: Enhancing Inverse Reinforcement Learning with Supervised Expert Guidance

Researchers introduce Hybrid-AIRL, an enhanced inverse reinforcement learning framework that combines adversarial learning with supervised expert guidance to improve reward function inference in complex, imperfect-information environments like poker. The method demonstrates superior sample efficiency and learning stability compared to traditional AIRL, particularly in settings with sparse and delayed rewards.

Commvault has a Ctrl+Z for rogue AI agents

Has Google’s AI watermarking system been reverse-engineered?

AI-generated synthetic neurons speed up brain mapping

Coming soon: 10 Things That Matter in AI Right Now

A publication is compiling its annual list of 10 breakthrough technologies expected to shape 2026, with a focus on AI, energy, and biotech. The article hints at potential challenges in selection this year, suggesting unprecedented difficulty in identifying which emerging technologies will have the most significant impact on society.

Bringing people together at AI for the Economy Forum

Google is hosting an AI for the Economy Forum in Washington D.C., bringing together stakeholders to discuss artificial intelligence's economic impact. The event represents a significant effort by major tech companies to shape policy dialogue around AI development and deployment.

Ontological Trajectory Forecasting via Finite Semigroup Iteration and Lie Algebra Approximation in Geopolitical Knowledge Graphs

Researchers introduce EL-DRUIN, an ontological reasoning system that uses finite semigroup algebra and Lie algebra to forecast geopolitical relationship trajectories rather than relying on LLM pattern matching. The system models political dynamics as composable states, identifies convergence points (attractors), and provides calibrated probability estimates for long-term geopolitical outcomes, with applications to scenarios like US-China technology decoupling.

ACE-TA: An Agentic Teaching Assistant for Grounded Q&A, Quiz Generation, and Code Tutoring

ACE-TA is an AI framework that combines large language models with three coordinated modules to provide automated educational support for programming students, including grounded question-answering, adaptive quiz generation, and interactive code tutoring with step-by-step guidance and sandboxed execution.

Real-Time Voicemail Detection in Telephony Audio Using Temporal Speech Activity Features

Researchers developed a lightweight machine learning system that detects voicemail greetings versus live human answers in real-time telephony audio with 96.1% accuracy using only temporal speech activity patterns. The system processes calls in 46ms on standard CPUs and has been validated across 77,000 production calls, achieving practical false positive and negative rates suitable for AI calling applications.

Wolkowicz-Styan Upper Bound on the Hessian Eigenspectrum for Cross-Entropy Loss in Nonlinear Smooth Neural Networks

Researchers derive a closed-form upper bound for the Hessian eigenspectrum of cross-entropy loss in smooth nonlinear neural networks using the Wolkowicz-Styan bound. This analytical approach avoids numerical computation and expresses loss sharpness as a function of network parameters, training sample orthogonality, and layer dimensions—advancing theoretical understanding of the relationship between loss geometry and generalization.

Product Review Based on Optimized Facial Expression Detection

Researchers propose a facial expression recognition system using a modified Harris algorithm to optimize product reviews by analyzing customer reactions in retail environments. The method reduces computational complexity while maintaining accuracy, enabling faster real-time detection of facial features for consumer sentiment analysis.

Enhancing Multimodal Large Language Models for Ancient Chinese Character Evolution Analysis via Glyph-Driven Fine-Tuning

Researchers have developed GEVO, a glyph-driven fine-tuning framework for multimodal large language models designed to analyze the evolution of ancient Chinese characters. The study introduces a comprehensive benchmark with 11 tasks and over 130,000 instances, demonstrating that even smaller 2B-scale models can achieve significant performance improvements in understanding character evolution and historical text transformation.

Enhanced-FQL($\lambda$), an Efficient and Interpretable RL with novel Fuzzy Eligibility Traces and Segmented Experience Replay

Researchers propose Enhanced-FQL(λ), a fuzzy reinforcement learning framework that combines fuzzified eligibility traces and segmented experience replay to improve interpretability and efficiency in continuous control tasks. The method demonstrates competitive performance with neural network approaches while maintaining computational simplicity through interpretable fuzzy rule bases rather than complex black-box architectures.

Controlling Multimodal Conversational Agents with Coverage-Enhanced Latent Actions

Researchers propose a novel reinforcement learning approach for fine-tuning multimodal conversational agents by learning a compact latent action space instead of operating directly on large text token spaces. The method combines paired image-text data with unpaired text-only data through a cross-modal projector trained with cycle consistency loss, demonstrating superior performance across multiple RL algorithms and conversation tasks.