Graph neural networks (GNNs) and graph transformers enable holistic modeling of clinical data, enhancing multimodal integration and interpretability.

Ruth Johnson, Michelle M. Li, Ayush Noori et al.

This study analyzes multi-head attention's optimization and generalization, demonstrating theoretical guarantees for training convergence and error bounds.

Puneesh Deora, Rouzbeh Ghaderi, Hossein Taheri et al.

The CAT framework employs environment augmentation and probabilistic decomposition for efficient closed-loop adversarial training, significantly improving autonomous driving safety.

Linrui Zhang, Zhenghao Peng, Quanyi Li et al.

Proposes Dual Bank Normalization (DBNORM) with DualIS and DualDIS to mitigate hubness, boosting cross-modal retrieval performance.

Yimu Wang, Xiangru Jian, Bo Xue

Proposes a novel k-NN based non-parametric conditional independence test for mixed continuous and categorical variables, improving robustness and accuracy.

Oana-Iuliana Popescu, Andreas Gerhardus, Jakob Runge

ReMax, based on REINFORCE, eliminates the value model, reduces GPU memory and training time, achieving SOTA for 7B models.

Ziniu Li, Tian Xu, Yushun Zhang et al.

MatFormer enables elastic inference via nested Transformer, supporting model extraction from 582M to 850M.

Devvrit, Sneha Kudugunta, Aditya Kusupati et al.

This study analyzes RLHF's impact on LLM generalization and diversity, showing improved out-of-distribution performance but reduced output variety.

Robert Kirk, Ishita Mediratta, Christoforos Nalmpantis et al.

Proposes Few-Shot In-Context Attack (ICA) and Defense (ICD) methods, significantly influencing LLM safety with minimal demonstrations.

Zeming Wei, Yifei Wang, Ang Li et al.

Proposes a solution-generator framework using policy gradient with structured features and entropy regularization, ensuring near-optimal solutions for combinatorial problems.

Constantine Caramanis, Dimitris Fotakis, Alkis Kalavasis et al.

Proposes constrained RLHF to prevent reward model overoptimization by dynamically regulating component weights via Lagrange multipliers, improving evaluation stability.

Ted Moskovitz, Aaditya K. Singh, DJ Strouse et al.

This paper introduces MLAgentBench, a benchmark for evaluating language model-based agents on 13 machine learning tasks, with Claude v3 Opus achieving an average success rate of 37.5%.

Qian Huang, Jian Vora, Percy Liang et al.

Using linear probes, the study reveals that Llama-2 models learn multi-scale linear representations of space and time, with identifiable 'space' and 'time neurons'.

Wes Gurnee, Max Tegmark

Epidemic Learning (EL) uses randomized communication topologies to achieve faster convergence than static methods, with a theoretical transient iteration bound of O(n^3/s^2).

Martijn de Vos, Sadegh Farhadkhani, Rachid Guerraoui et al.

Using GPT-4 to generate scientific paper feedback, achieving 30-40% overlap with human reviewers across 15 journals and ICLR, comparable to inter-reviewer agreement.

Weixin Liang, Yuhui Zhang, Hancheng Cao et al.

DeltaXplainer dynamically explains model differences via decision rules, enhancing model selection and monitoring efficiency.

Adam Rida, Marie-Jeanne Lesot, Xavier Renard et al.

Systematic analysis of activation patching metrics and methods; STR, logit difference, and sliding window outperform alternatives.

Fred Zhang, Neel Nanda

LogGPT leverages GPT-2 with reinforcement learning for log anomaly detection, outperforming SOTA with F1 scores over 0.9 on multiple datasets.

Xiao Han, Shuhan Yuan, Mohamed Trabelsi

This paper demonstrates exponential computational advantages of multimodal learning over unimodal, based on a geometric intersection problem reformulated via a special transformation.

Zhou Lu

Proposes a theoretical framework for multimodal learning, showing bounds up to O(√n) under connection and heterogeneity conditions.

Zhou Lu