833 papers
This paper demonstrates that the widely cited low-dimensional nature of failure subspaces in test-time adaptation is an algebraic artifact of specific gradient definitions constrained by the classifier's row space, rather than an intrinsic property, as alternative unconstrained demands reveal high-dimensional, width-sensitive structures where most energy remains confined to the classifier's row space.
This paper presents a systematic optimization study on the AMD Zen 3 microarchitecture that achieves 85.30 GFLOPS in single-core FP32 matrix multiplication by evaluating 28 distinct configurations of cache/register blocking, FMA chaining, and on-the-fly packing, ultimately identifying a champion design that reaches 63.5% of the theoretical peak performance while introducing a predictive model for future optimizations.
This paper proposes a multimodal deep learning framework that integrates textual, visual, and behavioral data from social media to achieve more accurate and early detection of mental health disorders compared to traditional machine learning and unimodal approaches.
This paper proposes a cost-effective method for simulating large LLM-agent societies on a laptop by replacing individual agents with low-parameter surrogates fitted from cheap queries, introducing a perception-memory taxonomy to predict simulation accuracy and validating the approach across multiple macroscopic scenarios.
This paper benchmarks FP16 GEMM performance on NVIDIA's Ada Lovelace RTX 4060 across cuBLAS, CUTLASS, and custom WMMA implementations, revealing that a pipeline depth of three with specific tile geometry achieves 52.7% of peak throughput while demonstrating that shared-memory pressure, rather than pipeline depth, is the primary bottleneck limiting further optimization.
This paper presents a multi-agent AI framework that integrates rule-based analysis, public RAG, Groq-powered LLM reasoning, and Q-learning reinforcement learning to autonomously detect, refactor, and validate code quality improvements in software repositories, achieving significant reductions in technical debt while preserving functional correctness.
This paper proposes a "simulated morphic field"—a computational metaphor for a persistent shared-memory layer—to enhance the efficiency and sustainability of multi-agent robotic learning while simultaneously serving as a pedagogical tool for teaching collective intelligence and systems thinking in engineering education.
This paper introduces PromptCanary, an open-source tool that applies golden-master regression testing to detect silent behavioral drift in Large Language Model APIs by comparing current outputs against versioned baselines using customizable prompts and scoring probes.
This paper presents ANA, an AI-driven multimodal application that integrates text, speech, and video analysis with Internal Family Systems (IFS) therapy principles to accurately recognize users' emotional states and internal psychological parts, achieving high performance across modalities to enable richer, embodied mental health interactions.
This paper proposes a methodology that combines shapelet-based time series analysis with multidimensional data structures and range constraints to precisely identify the temporal windows and user connectivity patterns driving hate speech propagation on social media, thereby enabling more effective event detection and mitigation.