GC-content bias in microarray and RNA-seq platforms is well-documented but rarely corrected in differential expression analyses. We audit 20 widely-cited microarray datasets from GEO, applying a permutation-based test that evaluates whether the overlap between differentially expressed gene lists and GC-content-correlated genes exceeds chance.
SNNs promise energy efficiency via sparse spike trains, but accuracy requires sufficient timesteps, creating a latency-accuracy tradeoff. We characterize this for 8 SNN architectures on CIFAR-10/100 and DVS-Gesture at timesteps 1-128.
The sim-to-real transfer gap is assumed to grow with task complexity, but we find a U-shaped relationship. Across 6 manipulation tasks (reaching, pushing, pick-and-place, stacking, insertion, bimanual assembly) with 5 domain randomization levels on Franka Emika: simple tasks transfer well (gap 8-12%), moderate tasks show maximum gap (28-41%), complex tasks show reduced gap (18-24%).
In cooperative MARL, free-riding agents contribute minimally while benefiting from team rewards. We propose Shapley Contribution Tracking (SCT) using online Shapley value approximation.
Multi-agent LLM systems chain multiple model instances via natural language, but scaling properties are unknown. We study 2-16 agents across four patterns (sequential, broadcast, hierarchical, peer-to-peer).
Fault-tolerant LLM training requires periodic checkpointing. We analyze the cost structure across 64-4,096 GPUs, comparing checkpoint overhead against failure recovery cost.
Distributed LLM training suffers from straggler nodes that impose synchronization barriers. We analyze 2,400 training runs on clusters of 10-512 GPUs with data/tensor/pipeline parallelism.
LLM APIs process inputs autoregressively, coupling response latency to input/output length. We demonstrate this creates an exploitable timing side channel: observing only response time reveals input token count with 93.
Prompt injection is a critical LLM security vulnerability. We analyze the tradeoff between injection resistance and helpfulness across 12 models from 4 families.
Technical debt density distribution across projects is poorly understood. We analyze 8,247 projects (6 languages) via SonarQube.
LLMs generate unit tests with impressive coverage, but we challenge this optimism using mutation testing. We evaluate GPT-4, Claude-3, CodeLlama-34B, and DeepSeek-Coder-33B on 200 Python functions from popular libraries.
Code review thoroughness is believed to decrease with PR size, but quantitative evidence is scarce. We analyze 50,247 reviews from 187 open-source GitHub repositories.
Semantic segmentation quality measured by IoU treats all pixels equally, but boundary pixels are inherently ambiguous and annotator agreement drops to near-chance there. We propose Attention Map Entropy (AME) computed from self-attention maps at the penultimate layer of ViT-based segmentation models.
Small object detection remains challenging despite architectural advances. We characterize resolution dependence by evaluating 6 detectors (YOLOv8, DETR, Faster R-CNN, DINO, Co-DETR, RT-DETR) on VisDrone and DOTA at 8 resolutions from 320×320 to 2560×2560.
Vision Transformers were hypothesized to be more shape-biased than CNNs due to global attention, but findings are contradictory. We resolve this through Fourier-domain selective masking: removing spatial frequency bands from ImageNet images and measuring accuracy degradation.
Learning rate warmup is near-universal in deep learning training, yet the optimal warmup duration is typically found through expensive grid search. We conduct a controlled comparison across Transformers and State-Space Models (Mamba) on language modeling, image classification, and time-series forecasting, training 840 models with warmup durations from 0 to 20% of training.
Feature attribution methods—Integrated Gradients, SHAP, LIME, Attention, GradCAM—often disagree on the same input. We investigate whether this disagreement is systematic by measuring pairwise agreement (Kendall's τ and top-k overlap) as a function of model depth.
Grokking—sudden generalization long after memorization—is difficult to predict. We identify a precursor: the Gradient Acceleration Index (GAI), the second derivative of gradient norm w.
Learning rate warmup is near-universal in deep learning training, yet the optimal warmup duration is typically found through expensive grid search. We conduct a controlled comparison across Transformers and State-Space Models (Mamba) on language modeling, image classification, and time-series forecasting, training 840 models with warmup durations from 0 to 20% of training.
The double descent phenomenon—where test error first decreases, then increases, then decreases again as model complexity grows—has been extensively documented under in-distribution evaluation. We investigate whether double descent persists under distribution shift by training 2,100 models (7 architectures × 6 widths × 50 seeds) on CIFAR-10 and evaluating under five controlled shift types: covariate shift (Gaussian noise), label shift (10% flip), domain shift (CIFAR-10.