Bilateral Decoupled Decay improves soft clipping in LLM reasoning with RLVR, addressing gradient divergence and enabling better exploration during policy optimization.
ES-Merging merges biological multimodal LLMs using embedding space signals to enable cross-modal scientific discovery beyond single-modality specialization.
OFA-TAD proposes generalist one-for-all anomaly detection for tabular data with cross-domain generalization, replacing dataset-specific training approaches.
Trainless GUI grounding for MLLM-based agents using element-level inference to map natural language to UI components without fine-tuning or large datasets.
Combines causal representation learning with local sparse attention for system identification, enabling interpretable deep learning of dynamical systems without predefined function libraries.
Unlearning-based sliding window approach for continual learning under concept drift, enabling models to adapt to non-stationary data streams without explicit task boundaries.
Proposes trust-region search algorithm for aligning diffusion and flow models to target rewards at inference time without requiring differentiable reward models.
Demonstrates deep neural networks can meta-learn task sequencing from few demonstrations, enabling generalization to new sequencing problems without task-specific training.
CausalEvolve improves LLM-based AI agents for open-ended scientific discovery by adding causal guidance and knowledge organization mechanisms to program evolution.
Extends critic match loss landscape visualization from online to off-policy reinforcement learning to reveal optimization geometry in critic learning.
FlashHead provides efficient drop-in replacement for classification head in language model inference, reducing parameters and compute by ~50%.
Delightful policy gradient method that addresses variance issues in policy gradient updates by accounting for action likelihood under current policy.
Proactive routing system that selects between black-box models and interpretable surrogates with distribution-free safety guarantees.
EARCP ensemble architecture dynamically weights heterogeneous expert models based on performance and inter-model coherence for sequential decision making.
AgentTrace provides causal graph tracing for post-hoc failure diagnosis in deployed multi-agent systems through execution log analysis.
Cross-RAG applies retrieval-augmented generation with cross-attention to improve zero-shot time series forecasting using foundation models.
DeFRiS applies decentralized federated reinforcement learning for IoT application scheduling across heterogeneous devices while preserving privacy.
GNNVerifier uses graph neural networks to verify and correct task plans generated by LLMs in autonomous agent systems, reducing hallucinations.
CAMD proposes coverage-aware decoding for multimodal LLMs to allocate compute efficiently by identifying easy vs hard reasoning cases.
POLCA framework uses LLMs as optimizers to automatically improve complex systems like prompts and multi-turn agents through numerical rewards and text feedback.
HO-SFL proposes hybrid-order split federated learning to reduce memory costs of backpropagation on edge devices while maintaining convergence speed.
Universe Routing framework addressing epistemic control in self-evolving agents by managing epistemologically incompatible reasoning frameworks.
OpenReservoirComputing: Python library for GPU-accelerated reservoir computing in JAX with automatic differentiation and JIT compilation.
Theoretical analysis of dataset distillation showing how gradient-based learning extracts and encodes task-relevant information into synthetic data.
Mechanistic analysis of multi-stream transformer architectures with manifold-constrained hyper-connections using ablation and causal methods.
Sample-efficient hypergradient estimation method for decentralized bi-level reinforcement learning with leader-follower agents.
Post-hoc explanation method using informative perturbation selection for model-agnostic ML explanations with uncertainty quantification.
Introduces directional routing mechanism for transformer attention heads with learned suppression directions, analyzed via mechanistic interpretability.
Proposes using LLMs as graph kernels for learning on text-rich graphs, treating text dynamically in message passing instead of static embeddings.
Heterogeneous spiking federated learning framework using fire-rate fusion for resource-constrained clients with SNNs.
Lightweight personalization method for split computing inference on edge devices handling distribution shifts and communication unreliability.
Log-barrier regularization improves exploration in Stochastic Gradient Bandit algorithm for policy optimization with global convergence guarantees.
MONET framework models neural network training efficiency from edge to data centers, capturing memory and backpropagation constraints.
Machine unlearning approach designs models with key deletion mechanism to erase training sample influence without full training data access.
Muon optimizer enforces orthogonality via Stiefel manifold projection for stable neural network training under heavy-tailed noise conditions.
System for encrypted skill sharing between AI agents using AES-256-GCM encryption over XMTP protocol.
Google AI Studio adds Project Spend Caps and revised Usage Tiers for controlling Gemini API monthly costs.
Interview discussing enterprises struggling to implement AI with authentic use cases and faking adoption.
Six AI agents for Claude Code that run locally in markdown files with no external dependencies, platform, or data collection.
Context Hub provides versioned, curated API documentation for coding agents to reduce hallucination and improve learning across sessions.
ssh.bot provides controlled SSH access for AI agents with granular permissions, audit trails, and kill-switch controls.
Stream0 messaging infrastructure for multi-agent communication with persistent inboxes and mid-task conversations.
LLM-powered web browsing tool with customizable interface. Limited details provided.
Monitoring platform tracking AI product quality across models and endpoints with real-time user experience metrics.
PostgreSQL extension enabling TypeScript function writing via Deno runtime with Node.js API support. Alpha quality.
Local AI inference platform replacing online stack with open-weight models, FLUX, and alternative LLM services.
Framework describing four levels of AI-driven engineering adoption, from code assistance to autonomous multi-step agent systems.
AI agent autonomously improved OWASP CRS regex detection rules: TPR 55.8%→100%, FPR 29.7%→4.8% across 20 experiments with 0 rejections.
GitHub Copilot adds Model Context Protocol support enabling persistent memory and external tool integration in Agent Mode.
Shhh: tool masking PII in AI prompts by replacing secrets with realistic fakes while preserving data structure for model reasoning.