Method extracting directional curvature information from Armijo backtracking line search as low-cost sharpness probe for adaptive learning rate safeguards.
Study showing reinforcement learning for vision-language model adaptation exhibits catastrophic forgetting, challenging assumptions about RL continual learning.
Theoretical analysis establishing minimum block width requirements for universal approximation in residual neural networks with inner width one.
Layer-parallel inference technique for encrypted Transformer computation using structured Newton methods to reduce FHE bottlenecks.
Training-free acceleration method for diffusion and flow matching models using x-prediction to reduce neural function evaluations during sampling.
Meta-learning approach to train neural network optimizers that outperform hand-designed optimizers like Adam, improving long-horizon training efficiency.
Learnable spectral positional encodings for directed graphs using Hermitian block Krylov subspaces, addressing eigendecomposition efficiency.
Bayesian decision-making framework for experimental design under adversarial uncertainty, focusing on reliable downstream decision-making.
HiQA proposes hierarchical contextual augmentation for retrieval-augmented generation in multi-document question-answering, improving accuracy and reducing hallucinations in LLM responses.
Theoretical analysis of two-stage training dynamics in transformers showing progression from syntactic to semantic feature learning.
Pipeline detecting spurious visual cues in multimodal LLMs using GPT-4 and object detectors without human supervision.
Metric and benchmark for measuring AI ability to complete long software tasks by correlating model performance to human completion time.
Network pruning method modeling pruning as flux-driven system with theoretical analysis of weight removal importance.
Study evaluating whether LLMs understand code semantics or rely on pattern matching, testing 10 state-of-the-art models on long context code understanding.
Analysis arguing AI productivity gains are primarily in developer tools and coding assistance.
Claude Science workbench for researchers with LLM capabilities, MCP connections, and scientific tools for discovery acceleration.
MCP tool enabling AI agents to reverse engineer binary code and understand feature implementations without source code.
Announcement of GLM-5.2 model availability on Canopy Wave platform with free trial offers.
ICML keynote slides on adapting work as AI capabilities increase. Discussion of future implications without technical research.
Project using LLM jury systems to build food metadata. Limited information provided about methodology or implementation.
News about SFU professor and legal AI startup researching searchable court decisions for self-represented litigants. Application domain announcement.
Open source LLM router that finds cheaper inference across multiple providers within specified time/cost budgets across 62,852 requests.
Article about designing LLMs with hardware-friendly architecture co-design principles.
Developer tool providing authentication and verification for AI voice agents, with decorators for protecting sensitive operations.
Daftari long-term memory system for AI agents that maintains relational context and tensions across past interactions like a historical ledger.
Screen-aware AI tutor (HeyBraza) that provides real-time guidance on desktop applications via natural language interaction.
AI agent (FixBugs) that reproduces production bugs in sandboxes and generates verified fixes, available as VSCode extension and GitHub app.
System extending Cursor IDE to maintain longer context history and session continuity across 20+ minute intervals.
Deterministic context compaction for AI agents using BPE tokenization instead of LLM summarization, achieving 92.6% cache hit rate and 63-72% cost reduction.
Security analysis tool for AI-generated code that tracks supply chain vulnerabilities, showing AI assistants introduce 10x more security findings than humans.
Open source package manager (sx 2.0) for sharing AI skills, MCP configs, and prompts across teams without git/terminal requirements.
Shared memory layer tool for AI assistants and teams, centralizing scattered context (prompts, conventions, examples) across Claude, ChatGPT, and markdown.
Essay on designing programming languages optimized for LLM interaction by prioritizing human-readable syntax and training distribution alignment.
Opinion piece on leveraging current AI subsidies from frontier labs to improve open source infrastructure and sustainability.
Browser-based IDE with AI agents that build applications from natural language descriptions, integrating external services via API context.
AI coding agent that automatically blocks commits failing security scans, preventing vulnerable code deployment.
Portable context layer (Forgein) using MCP servers to sync AI context across machines, sessions, and tools without repetition.
Discussion on containerization and security best practices for AI agents accessing local systems and OS files. Technical security guidelines sought.
Reproducible evaluation harness demonstrating agent-eval vulnerabilities and cheating via test environment tampering, proposing per-task isolation fixes.
Claude Code plugin adding Mr. Meeseeks sound notifications when Claude awaits user input, with filtering for autonomous vs. interactive work.
Speculates on LLM internal representations across layers, proposing middle layers might represent 'subconsciousness'. Article appears incomplete.
Empirical study analyzing characteristics and quality of AI-generated code in open source repositories.
MindRoom platform for deploying AI agents across environments.
MIT method for detecting AI models trained on assembly code without generating it.
Open source PostgreSQL workload generator tool for stress testing and database performance evaluation, with warnings for testing-only use.
Overview of AI agents deployed for revenue generation across organizations.
Lil'Log article exploring recursive self-improvement in AI systems, tracing the concept from Good (1965) through modern AI feedback loops.
Fleet Deck: Open-source dashboard for managing multiple parallel Claude Code sessions with status tracking and queue visualization.
MIT research on replacing LLM inference with deterministic code-based checks for standards validation.
Meta Muse Spark 1.1 LLM benchmark results showing 8-point improvement over 1.0 with cost and token efficiency metrics.