AI & LLMs
The deepest channel on Edge. Foundation models, agent architectures, retrieval, evals, and the moving line between research and production.
GLARE: Natural Language Interfaces for Global Model Explanations
GLARE provides a natural language interface for querying global model explanations, allowing users to interpret complex AI behavior through conversational prompts rather than static visualizations.
Moving Beyond Static Leaderboards for LLM Agent Evaluation
Static benchmarks often fail to predict real-world performance for LLM agents; the authors propose a framework focused on predictive validity to better align evaluation with practical utility.
Toten: Ontological Tokenization for Technical Portuguese
Toten is a knowledge-based tokenization framework designed to accurately parse physical quantities and technical notation in Brazilian Portuguese, addressing common failures in standard NLP tokenizers.
The Symbiotic Evolution of AI and Software Engineering
The intersection of AI and Software Engineering (AI4SE and SE4AI) has matured over the last decade, shifting from experimental research to essential production-grade methodologies for building, testing, and maintaining complex systems.
Configurable Clinical Information Extraction with Agentic RAG
Agentic RAG systems for clinical data require modular configuration to balance precision and recall, as monolithic pipelines often fail to handle the high variability of medical documentation.
Optimizing LLM Post-Training Through Pairwise Comparison Selection
The paper investigates how the selection of response pairs in preference-based post-training (like DPO or PPO) impacts model performance, suggesting that strategic pair selection is as critical as the training algorithm itself.
Detecting LLM Epistemic Blind Spots via Cross-Model Attribution
LLMs often hallucinate confidence in clinical settings. This paper introduces a method using Cross-Model Attribution Divergence (CMAD) to identify when models rely on unreliable features, effectively flagging epistemic uncertainty in tabular data.
Deontic Policies for Runtime Governance of Agentic AI
The paper proposes using deontic logic—a system of formal rules defining obligations, permissions, and prohibitions—to govern the runtime behavior of autonomous AI agents.
Building Reliable AI Code Generation Pipelines with Salesforce CodeGen
To move AI-generated code from prototype to production, implement a multi-stage pipeline that includes automated unit testing, safety sandboxing, and model-based reranking to filter out hallucinated or insecure outputs.
Liquid AI's New 350M Multilingual Retrieval Models
Liquid AI has released LFM2.5-Embedding-350M and LFM2.5-ColBERT-350M, two efficient, bidirectional retrieval models optimized for multilingual search across 11 languages.
Perplexity Brain: Self-Improving Memory for AI Agents
Perplexity's 'Brain' system shifts AI memory from user-centric profiles to agent-centric performance, using an overnight context graph to learn from past tasks, failures, and corrections to improve future efficiency.
The Rise of Agentic Traffic and Microsoft's Model Strategy
Agentic AI bots now dominate web traffic, signaling a shift in how we interact with information. Meanwhile, Microsoft is pivoting to first-party models, prioritizing safety and cost-efficiency for enterprise users.
Architecting Long-Running AI Agents for Multi-Day Workflows
Move beyond stateless chatbots by implementing event-driven dormancy, durable checkpointing, and decoupled evaluation to manage complex, multi-day workflows.
Google Cloud TechBuilding AI Agents with Model Context Protocol (MCP)
The Model Context Protocol (MCP) acts as a universal adapter, allowing AI agents to securely interact with external tools and live data via a standardized input/output interface, decoupling agent logic from tool implementation.
The Production AI Playbook: Deploying Agents at Enterprise Scale
Moving AI from demo to production requires shifting focus from model selection to five pillars: evaluation, observability, data foundation, orchestration, and governance.
RODS: Improving Multi-Turn Tool-Use Agents via Reward-Driven Synthesis
RODS (Reward-Driven Online Data Synthesis) improves multi-turn tool-use agents by generating high-quality synthetic training data through iterative reward-based filtering, addressing the scarcity of complex, multi-step interaction data.
Skill-Guided Continuation Distillation for GUI Agents
The paper introduces a method to improve GUI agent performance by distilling complex task trajectories into modular, skill-based sub-tasks, enhancing generalization and execution reliability.
Decoupling Search from Reasoning in LLM Agents
Native search grounding in LLMs creates rigid, expensive, and opaque agent architectures. Moving to a Decoupled Search Grounding (DSG) layer allows for vendor-agnostic control over retrieval, caching, and cost, while maintaining accuracy.
SciRisk-Bench: Evaluating Safety in AI for Science
SciRisk-Bench is a new benchmark designed to evaluate the safety risks of AI models specifically applied to scientific research, focusing on multi-dimensional risk assessment.
Improving AI Scientist Reliability via Research Harnesses
The paper proposes a 'Research Harness' to externalize synthesis and validation, addressing the reliability issues inherent in autonomous AI research agents.
WorldLines: Benchmarking Long-Horizon Stateful Embodied Agents
WorldLines introduces a new benchmark and modeling framework designed to evaluate how embodied AI agents maintain state and execute complex, long-horizon tasks over extended periods.
DeFAb: A New Benchmark for Defeasible Abduction in LLMs
DeFAb is a new, verifiable benchmark designed to test how well foundation models handle defeasible abduction—the ability to form logical explanations that can be retracted or revised in light of new, contradictory information.
CEO-Bench: Measuring Long-Term Strategic Reasoning in AI Agents
CEO-Bench is a new evaluation framework designed to test whether AI agents can maintain strategic coherence and decision-making over extended, multi-step business scenarios.
Vercel's Eve: A Filesystem-First Framework for AI Agents
Vercel has released Eve, an open-source framework that treats AI agents as directories of files, mapping specific capabilities like tools, skills, and schedules to file paths to eliminate boilerplate and production plumbing.
The KV Cache Compression Race: TurboQuant vs OSCAR vs EpiCache
KV cache compression is the new frontier for scaling LLM inference, with TurboQuant, OSCAR, and EpiCache offering distinct strategies to balance memory footprint against model accuracy.
The Shift Toward User-Controlled AI Recommendation Algorithms
Major social platforms are moving from opaque, one-size-fits-all algorithms to user-tunable systems, leveraging LLMs to allow granular control over feed content.
Building AI Agents with Google's Agent Development Kit (ADK)
A practical walkthrough on using Google's Agent Development Kit (ADK) to build autonomous agents that can interact with text-based environments, specifically demonstrated through a retro-inspired adventure game.
Solving the Physical AI Data Bottleneck
XDOF is building the infrastructure for physical AI by providing the high-fidelity, large-scale training data that robotics models currently lack, moving beyond the limitations of low-quality video data.
Pramaana Labs Uses Formal Verification to Secure Enterprise AI
Pramaana Labs raised $27M to integrate formal verification—using the LEAN programming language—with LLMs to ensure deterministic, error-free outputs in high-stakes fields like tax, law, and drug discovery.
Predicting AI Model Behavior via Deployment Simulation
OpenAI uses 'Deployment Simulation'—replaying real, de-identified user conversations with new models—to predict safety risks and undesired behaviors before public release, outperforming traditional synthetic evaluations.
Showing 30 of 758