Models, papers, tools. 109,441 articles with AI-powered sentiment analysis and key takeaways.
OpenSkill introduces a framework enabling LLM agents to self-evolve in open-world environments without task-specific supervision, bootstrapping both skills and verification signals from public documentation and web resources. The approach demonstrates superior performance across benchmarks while maintaining transferability across different models, addressing a critical gap in autonomous agent deployment.
A comprehensive survey examines latent space as an emerging computational substrate for language models, arguing that continuous latent representations are more efficient than explicit token-level generation for critical internal processes. The research identifies four mechanistic developments (architecture, representation, computation, optimization) and seven capability areas (reasoning, planning, modeling, perception, memory, collaboration, embodiment) that latent space enables.
Researchers introduce EVA, an evolutionary framework that demonstrates GUI agents powered by multimodal language models are vulnerable to Environmental Injection Attacks through semantic deception rather than visual manipulation, achieving 85% attack success rates and revealing a critical security flaw in instruction-following alignment training.
Researchers demonstrate that Whisper, OpenAI's widely-used speech recognition model, can detect and mitigate hallucinations—fabricated coherent transcriptions from non-speech audio—using Sparse AutoEncoders and activation-space steering. The approach reduces hallucination rates from 72-87% to 14-27% across model sizes with minimal performance degradation on actual speech.
A comprehensive study of 8 state-of-the-art language models reveals significant limitations in probabilistic reasoning, with accuracy dropping from 96% on standard problems to 59% on counterintuitive ones. The research demonstrates that LLMs are vulnerable to token bias and prompt manipulation, suggesting they lack genuine probability reasoning despite excelling at other mathematical tasks.
Researchers introduce MemDreamer, a framework that enables Vision-Language Models to process hours-long videos by decoupling perception from reasoning through hierarchical graph memory and agentic retrieval. The approach achieves state-of-the-art results while reducing computational context requirements to 2% of full video ingestion, establishing a new paradigm for long-form multimodal understanding.
Researchers introduce Insights Generator (IG), a multi-agent system that automates the diagnosis of failures in large language model agents by analyzing execution trace corpora at scale. IG produces evidence-backed natural language insights about systematic behavioral patterns, demonstrating 30.4 percentage point performance improvements when human experts implement its recommendations.
Researchers conducted an empirical comparison of mathematical reasoning between humans and DeepSeek-R1, analyzing 10,247 reasoning steps across 30 AIME problems. The study reveals that while the AI model exhibits surface-level reasoning patterns, it engages in inefficient verification loops and lacks the structured deduction humans employ, suggesting current long-chain-of-thought models may be optimized for appearing to reason rather than reasoning effectively.
Researchers introduce CultureScore, a new evaluation framework for assessing cultural faithfulness in video generation models, revealing that leading AI systems like Veo 3.1 and LTX-2 fail to accurately represent diverse global cultures. Testing across 10 countries shows the best model achieves only 56.8% cultural accuracy, with human evaluators valuing cultural representation over visual quality metrics.
Socratic-SWE introduces a self-evolving framework that improves LLM-driven software engineering agents by distilling their solving traces into structured skills that guide targeted task generation. The approach achieves 50.40% on SWE-bench Verified after three iterations, demonstrating that agent weaknesses can fuel scalable, execution-validated training data creation without manual intervention.
Researchers propose formalizing the evaluation of foundation model agents through a classical sim-to-real framework based on Markov Decision Processes, addressing the gap between simulated training and real-world deployment. The work advocates adopting established robotics solutions like domain randomization and establishing standardized benchmarks to build more reliable AI agents for production applications.
Researchers demonstrate that Rectified Flows, a generative model architecture increasingly deployed in production systems, leak membership information about training data along their interpolation path in a quantifiable, bell-shaped pattern. This vulnerability enables practical membership inference attacks that can distinguish training set members from non-members, raising significant privacy and copyright concerns for deployed generative AI systems.
Researchers present a qualitative model for understanding AI sovereignty—how nations independently control AI technologies—identifying critical leverage points like electricity, data, and skilled talent that determine competitive advantage. The framework highlights both kinetic and non-kinetic methods nations may employ to strengthen their AI capabilities or undermine rivals, positioning AI as a central instrument of 21st-century national power competition.
Researchers have developed a new method called Controlled Latent-space Evasion that can bypass safety guardrails in language models by manipulating their internal representations more effectively than previous techniques. The attack reframes refusal suppression as an evasion problem against linear probes and achieves state-of-the-art success rates across 15 different models, highlighting a significant vulnerability in current AI safety alignment approaches.
Researchers propose COMPACT-VA, a planning-aligned token compression framework using conditional VQ-VAE to enable vision-action models in autonomous driving to process extended temporal context within real-time computational budgets. The approach achieves over 6% improvement in driving success rates while delivering 3.3x speedup and 2.7x memory reduction compared to uncompressed processing.
A research paper proposes the Three-Ring Architecture as a governance framework for enterprise AI deployment, arguing that organizations deploying agentic AI systems lack adequate control infrastructure. The framework separates deterministic, strategies-based agents (Ring 2) from non-deterministic LLM-based agents (Ring 3), positioning Ring 2 as essential operating system-level governance to prevent the 95% project failure rates seen in previous AI deployment waves.
Researchers measured how well frontier AI models perform complex reasoning without explicit chain-of-thought (CoT) tokens, finding that no-CoT task-completion time horizons have doubled yearly over six years. GPT-5.5 now reaches over 3 minutes of reasoning complexity, with projections suggesting frontier models could exceed 7 minutes by 2028 and 25 minutes by 2030, raising concerns about the effectiveness of current AI safety monitoring approaches.
Researchers identify a critical vulnerability in agent-interoperability protocols like A2A and MCP: while message content is encrypted, the communication metadata revealing which agents contact each other, when, and how often exposes pending workflows and enables adversaries to predict and preempt autonomous actions. The study demonstrates that observers can infer task classes from metadata patterns alone and that metadata-protecting transports significantly reduce this predictive leverage.
Researchers have introduced DuMate-DeepResearch, a multi-agent AI system designed to handle complex research tasks with improved auditability and reasoning. The framework achieves state-of-the-art results on deep research benchmarks by combining dynamic planning, recursive task delegation, and rubric-based quality optimization.
Native3D introduces an end-to-end 3D scene generation framework that eliminates the need for 2D intermediate representations, using a unified mesh-texture modeling approach with semantic alignment to improve geometric and textural fidelity compared to traditional diffusion model-based methods.
A study of Perplexity's autonomous AI agents reveals they perform 26 minutes of productive work per session versus 33 seconds for traditional search, reducing task completion time by 87% while improving quality and expanding the scope of work users attempt. This research demonstrates how AI agents are transitioning from conversational tools to end-to-end task executors that fundamentally reshape knowledge work.
Researchers introduce Zero-Shot Embedding Drift Detection (ZEDD), a lightweight defense mechanism that detects prompt injection attacks on large language models by measuring semantic shifts in embedding space. The method achieves over 93% accuracy with less than 3% false positives across multiple LLM architectures without requiring model access or task-specific training.
Researchers introduce STREAM, a novel framework applying Riemannian flow matching to synthetic histopathology image generation. The approach leverages pretrained Vision Foundation Models as latent space rather than conditioning signals, addressing the "conditioning collapse" problem and achieving state-of-the-art results for medical image synthesis.
OffQ introduces a novel quantization technique for large language models that addresses activation outliers through an offsetting mechanism, enabling efficient W4A4KV4 low-bit quantization. The method uses top-1 PCA to identify outlier subspaces and concentrates high-magnitude activations into a single channel via rotation, then converts this into a shared offset to reduce standard deviation. This approach maintains uniform-grid quantization while improving accuracy across diverse LLM architectures.
Researchers introduce SlimSearcher, a framework that trains AI web agents to perform complex information-seeking tasks with 17-58% fewer tool calls while maintaining or improving accuracy. The approach combines efficient trajectory filtering during supervised fine-tuning with adaptive reward gating during reinforcement learning to eliminate wasteful search behaviors.
