#llm
Every summary, chronological. Filter by category, tag, or source from the rail.
Steering LLM Personality via Latent Feature Interventions
Researchers have developed a mechanistic method to steer LLM personality traits by identifying and modifying latent features in the model's residual stream using sparse autoencoders, enabling precise behavioral control without retraining.
HyphaeDB: Moving From Passive Storage to Agent-Native Memory
HyphaeDB reinterprets HNSW graph topology as a communication fabric for multi-agent systems, enabling knowledge propagation and emergent consensus rather than just passive retrieval.
ComMem: Dual-Memory Systems for VLM Test-Time Adaptation
ComMem improves VLM robustness by mimicking biological memory, using a fast-adapting visual cache and a slow-integrating textual prototype system to maintain cross-modal consistency during test-time adaptation.
Agentic Abstention: Improving When LLM Agents Should Stop
LLM agents often fail to stop when a task is impossible, leading to unnecessary tool use. The CONVOLVE method improves timely abstention by distilling interaction trajectories into reusable stopping rules.
Agent Safety Is Action Alignment, Not Content Refusal
Treating agent safety like chatbot content moderation is a category error. True agent security requires enforcing least privilege at the action boundary, not training models to refuse requests.
Making LLM Self-Evolution Safe with Held-Out Selection
RSEA improves LLM agent performance by recursively evolving natural-language artifacts while using a strict held-out validation gate to prevent performance regression.
Stabilizing Critic-Free RL with BV-Blend
BV-Blend improves reinforcement learning stability by blending prompt-local statistics with historical cluster-based moments, preventing training stalls when reward variance is zero.
Closing the Loop Between Model Evaluation and Data Intervention
By introducing 'capability slices'—groups of evaluation samples categorized by task and operation—engineers can transform benchmark failures into precise, actionable data interventions rather than relying on intuition.
GPTNT: A Real-Time Collaborative Benchmark for AI Agents
GPTNT uses the game 'Keep Talking and Nobody Explodes' to test AI agent collaboration under time pressure, revealing critical failures in state tracking and real-time communication.
Optimizing LLM Inference: KV Cache and Paged Attention
LLM inference latency and throughput bottlenecks are often caused by inefficient GPU memory management. Using KV caching, paged attention, and specific tuning techniques like chunked prefill can drastically improve performance.
Why Vibe Coding Platform Base44 is Building Its Own AI Model
Base44 is transitioning to a vertically integrated stack by training its own LLM to gain control over latency, costs, and performance, signaling a shift toward defensibility for AI-native startups.
Stop Blaming Your RAG Pipeline: 16 Production Techniques
Most RAG failures are pipeline issues, not model limitations. Improving retrieval precision through hybrid search, reranking, and rigorous evaluation is more effective than simply swapping models.
Ornith-1.0: Coding Models That Learn Their Own Harness
Ornith-1.0 achieves state-of-the-art performance for its size by incorporating the coding harness into the model's training gradient, allowing the model to dynamically generate its own execution scaffolds rather than relying on static, human-written ones.
Optimizing RAG Retrieval with Hierarchical Search
Hierarchical RAG improves precision and reduces computational costs by replacing flat, corpus-wide similarity searches with a two-stage process: document-level filtering followed by targeted chunk retrieval.
Building Great Agent Skills: The Missing Manual
To escape 'skill hell,' developers must treat agent skills as structured, maintainable code by optimizing triggers, minimizing context bloat, using 'leading words' for steering, and aggressively pruning irrelevant instructions.
Building Production-Grade Multi-Agent Systems with ADK
Learn to build robust, state-aware multi-agent systems using Google's Agent Development Kit (ADK) and the Model Context Protocol (MCP) to handle orchestration, security, and persistence.
Scaling Item Knowledge with JD's Oxygen AIIC Platform
JD.com's Oxygen AIIC uses a hybrid LLM/VLM architecture to automate item-knowledge production at scale, achieving 94.2% precision and 82.8% recall across tens of billions of SKUs.
Agent-Native Immune System (ANIS): Architecture for Runtime Defense
The Agent-Native Immune System (ANIS) shifts AI security from static training-time alignment to dynamic, runtime defense, using a six-layer 'Immune Tower' to protect autonomous agents against memory poisoning and tool-chain manipulation.
DysLexLens: Analyzing Dyslexic AI User Experiences via LLMs
DysLexLens is an end-to-end framework that extracts, structures, and validates insights from noisy online forum data to understand how dyslexic learners interact with AI tools.
ToE: Hierarchical Claim Verification Against Adversarial Misinformation
Tree of Evidence (ToE) is a fact-checking framework that uses a reinforcement learning-driven agent to decompose claims into hierarchical argument trees, significantly improving verification accuracy against adversarially poisoned inputs.
Improving Long-Horizon LLM Planning via Symbolic Feedback
This framework enhances LLM planning reliability by using a symbolic verifier to identify errors and provide corrective, interpretable instructions for iterative self-refinement.
Personality Prompting in Multi-Agent Teams: Task-Dependent Impact
Personality manipulation in LLM agents significantly alters communication style but only degrades task performance in open-ended or collaborative domains, while remaining largely neutral in structured coding tasks.
The Shift in Software Engineering: AI Agents and Production Risk
AI agents have fundamentally transformed software development in six months, enabling massive increases in code output. However, this shift risks quality and security when organizations prioritize AI adoption over core engineering rigor, as evidenced by recent high-profile outages.
Building and Auditing Local Coding Agents
A practical guide to setting up a local coding agent stack using Ollama and open-weight models, emphasizing performance benchmarking, secure auditing of agent harnesses, and the trade-offs of running local vs. proprietary infrastructure.
Claude Tag: Moving AI from Chat to Team-Based Delegation
Claude Tag shifts LLM interaction from synchronous chat to asynchronous, team-wide delegation within Slack, positioning Claude as a persistent, proactive coworker rather than a standalone tool.
OpenAI's GPT-5.6 Launch: Frontier Models as Managed Assets
OpenAI released the GPT-5.6 family (Sol, Terra, Luna) as a restricted, government-mediated preview, signaling a shift where release governance is now a core component of the model specification.
Internal AI Adoption & The Rise of Agentic Workflows
OpenAI reports massive internal token growth across all departments, signaling that agentic workflows—supported by review loops and persistent infrastructure—are moving from experimental to core production patterns.
ParallelKernelBench: Frontier LLMs Struggle with Multi-GPU Kernels
While LLMs excel at single-GPU kernel generation, they currently struggle with multi-GPU tasks where communication bottlenecks and complex rank coordination dominate performance.
Deploying vLLM Endpoints on Hugging Face Jobs
Hugging Face Jobs allows engineers to spin up private, OpenAI-compatible vLLM endpoints on demand using a single command, providing a pay-per-second alternative for testing and experimentation.
Tandem Reinforcement Learning: Aligning AI Reasoning with Humans
Tandem Reinforcement Learning (TRL) forces stronger models to co-generate reasoning with weaker models, resulting in more legible, robust, and human-compatible chains of thought without sacrificing performance.
Showing 30 of 1042