A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
Nathan A. Baker, Frank Alexander, T. Bremer, et al.
This workshop report outlines core technologies and basic research needs for scientific machine learning and AI, emphasizing a national strategy for advancing the field.
Arda Sevinc, Abdurrahman Gumus
AutoReason automatically generates chain-of-thought rationales by decomposing implicit queries into explicit questions, improving multi-step reasoning accuracy in weaker LLMs on StrategyQA and HotpotQA.
Daniel Armstrong, Xuan-Vu Nguyen, Octavian Susanu, et al.
SynthEx, an agentic LLM framework, plans retrosynthetic routes to complex natural products that expert chemists judge comparable to human-designed syntheses.
Unknown
This study aggregates empirical evidence from data strategy studies to assess the effects of model quantization, finding a weak negative impact.
Unknown
DiffFit proposes a simple parameter-efficient fine-tuning strategy for large diffusion models, enabling transferability with minimal trainable parameters.
Unknown
Introduces RAS, a training-free sampling strategy for Diffusion Transformers that dynamically adjusts token updates based on region-adaptive criteria.
Unknown
This paper investigates whether model editing harms the general abilities of large language models and proposes regularization as a mitigation strategy.
Awni Altabaa, John Lafferty
This paper proposes a variational objective to decompose a language model's response distribution into strategy-conditioned latent codes, preventing posterior collapse by focusing on high-surprisal tokens.
Venkata Naga Sai Vishnu Rohit Pulipaka, Anish Katta, Deva Rohit Reddy Peddireddy
This paper shows that a native C++ ONNX Runtime inference engine can score RLHF rewards faster than PyTorch eager mode and FastAPI on CPU, but torch.compile is faster on GPU, and batching strategy matters more than language or runtime choice.
Ioannis Maniadis Metaxas, Adrian Bulat, Alberto Baldrati, et al.
UltraViT is a latency-optimized vision encoder for LVLMs using a pyramidal architecture with heterogeneous spatial mixers and a two-stage generative pre-training strategy, achieving 1.7x speedup on-device.
Yu Wang, Shiwan Zhao, Zhihu Wang, et al.
SCoT improves LLM reasoning by first eliciting a problem-solving strategy before generating Chain-of-Thought steps, achieving significant gains on reasoning benchmarks.
Jordan Pollack, Alan Blair
This paper demonstrates that simple hillclimbing with relative fitness can train a neural network to play backgammon at a competitive level, challenging the necessity of temporal difference learning.