A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
André Altmann, Laura Toloşi, Oliver Sander, et al.
Introduces PIMP, a permutation-based method that corrects RandomForest feature importance bias by computing P-values for variable importance.
Unknown
Proposes a fast, scalable nonparametric method to quantify uncertainty in predictions from a single deterministic neural network.
Unknown
Investigates neural scaling laws for the ERM minimizer in the teacher-student setting beyond the random features regime, focusing on feature learning.
Unknown
RocketKV is a training-free two-stage KV cache compression method that accelerates long-context LLM inference by combining coarse-grain permanent eviction with fine-grain dynamic sparsity.
Unknown
The paper proposes hierarchical process reward models as symbolic vision learners, supervising intermediate reasoning steps in vision tasks.
Mohammad Tavakoli, Alireza Salemi, Carrie Ye, et al.
Introduces BEAM, a benchmark for long-term memory in LLMs with up to 10M tokens, and LIGHT, a memory framework inspired by human cognition that improves performance by 3.5%-12.69%.
Yizhou Fan, Luzhen Tang, Huixiao Le, et al.
A randomized experiment shows ChatGPT improves short-term writing performance but may induce metacognitive laziness, with no gains in intrinsic motivation or knowledge transfer.
Mike A. Merrill, Alexander G Shaw, Nicholas Carlini, et al.
Terminal-Bench 2.0 is a hard benchmark of 89 real-world-inspired CLI tasks where frontier agents score below 65%, with error analysis and public dataset/harness.
Ziheng Qin, Yaxin Lu, Zhangyang Atlas Wang, et al.
StateM is an agent-native runtime that improves long-horizon agent performance via harness scaling, reaching 95.3% raw accuracy on Terminal-Bench 2.1 without changing model weights.
Unknown
This paper discusses the potential high-level and long-term benefits of using multimodal large language models (M-LLMs) in health care, including their ability to integrate diverse data types for improved clinical decision-making.
Ling Yue, K. Bhandari, Ching-Yun Ko, et al.
This survey reviews methods for designing and optimizing LLM agent workflows, treating them as agentic computation graphs and organizing literature by when workflow structure is determined.
Reef Menaged, G. Lior, Shauli Ravfogel, et al.
This paper introduces agentic automata learning to evaluate how well tool-calling LLM agents can uncover hidden deterministic finite automata through interaction, finding that performance drops sharply with DFA size and that current agents are less r