Daily Arxiv

This page organizes papers related to artificial intelligence published around the world.
This page is summarized using Google Gemini and is operated on a non-profit basis.
The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.

A Survey on Cache Methods in Diffusion Models: Toward Efficient Multi-Modal Generation

Learning To Defer To A Population With Limited Demonstrations

Every Attention Matters: An Efficient Hybrid Architecture for Long-Context Reasoning

Local Guidance for Configuration-Based Multi-Agent Pathfinding

Diagnosing Representation Dynamics in NER Model Extension

GUIDE: Enhancing Gradient Inversion Attacks in Federated Learning with Denoising Models

The Parameterized Complexity of Computing the VC-Dimension

TabR1: Taming GRPO for tabular reasoning LLMs

Efficient Vision-Language-Action Models for Embodied Manipulation: A Systematic Survey

Improving Model Representation and Reducing KV Cache via Skip Connections with First Value Heads

HumanCM: One Step Human Motion Prediction

SPLite Hand: Sparsity-Aware Lightweight 3D Hand Pose Estimation

Bridging Symmetry and Robustness: On the Role of Equivariance in Enhancing Adversarial Robustness

MARIS: Marine Open-Vocabulary Instance Segmentation with Geometric Enhancement and Semantic Alignment

Towards Robust Zero-Shot Reinforcement Learning

VaultGemma: A Differentially Private Gemma Model

Feature Selection and Regularization in Multi-Class Classification: An Empirical Study of One-vs-Rest Logistic Regression with Gradient Descent Optimization and L1 Sparsity Constraints

Bayes or Heisenberg: Who(se) Rules?

Epistemic-aware Vision-Language Foundation Model for Fetal Ultrasound Interpretation

Phenome-Wide Multi-Omics Integration Uncovers Distinct Archetypes of Human Aging

Empirical Study on Robustness and Resilience in Cooperative Multi-Agent Reinforcement Learning

HoMer: Addressing Heterogeneities by Modeling Sequential and Set-wise Contexts for CTR Prediction

On the Fairness of Privacy Protection: Measuring and Mitigating the Disparity of Group Privacy Risks for Differentially Private Machine Learning

FlyLoRA: Boosting Task Decoupling and Parameter Efficiency via Implicit Rank-Wise Mixture-of-Experts

Stress-Testing Model Specs Reveals Character Differences among Language Models

Relational Transformer: Toward Zero-Shot Foundation Models for Relational Data

A New Digital Divide? Coder Worldviews, the Slop Economy, and Democracy in the Age of AI

Replacing Softmax Similarity with a Sharpened Angular Similarity: Theory and Practice of Scaling To Billion-Context Attention

TriQuest:An AI Copilot-Powered Platform for Interdisciplinary Curriculum Design

Constraint Satisfaction Approaches to Wordle: Novel Heuristics and Cross-Lexicon Validation

Addressing Pitfalls in the Evaluation of Uncertainty Estimation Methods for Natural Language Generation

DragFlow: Unleashing DiT Priors with Region Based Supervision for Drag Editing

FerretNet: Efficient Synthetic Image Detection via Local Pixel Dependencies

TianHui: A Domain-Specific Large Language Model for Diverse Traditional Chinese Medicine Scenarios

WolBanking77: Wolof Banking Speech Intent Classification Dataset

PersonaMatrix: A Recipe for Persona-Aware Evaluation of Legal Summarization

Floq: Training Critics via Flow-Matching for Scaling Compute in Value-Based RL

Toward a Metrology for Artificial Intelligence: Hidden-Rule Environments and Reinforcement Learning

Benchmarking GPT-5 for biomedical natural language processing

Sparse Autoencoder Neural Operators: Model Recovery in Function Spaces

Conformal Prediction for Time-series Forecasting with Change Points

Multi-Agent Reinforcement Learning for Task Offloading in Wireless Edge Networks

LFD: Layer Fused Decoding to Exploit External Knowledge in Retrieval-Augmented Generation

Memory Decoder: A Pretrained, Plug-and-Play Memory for Large Language Models

Position: The Current AI Conference Model is Unsustainable! Diagnosing the Crisis of Centralized AI Conference

DMSC: Dynamic Multi-Scale Coordination Framework for Time Series Forecasting

Bayesian Optimization of Process Parameters of a Sensor-Based Sorting System using Gaussian Processes as Surrogate Models

MCIF: Multimodal Crosslingual Instruction-Following Benchmark from Scientific Talks

Fine-Tuning Multilingual Language Models for Code Review: An Empirical Study on Industrial C# Projects

Compositional Coordination for Multi-Robot Teams with Large Language Models

Quantization-Aware Neuromorphic Architecture for Efficient Skin Disease Classification on Resource-Constrained Devices

Frequency-Dynamic Attention Modulation for Dense Prediction

Crafting Imperceptible On-Manifold Adversarial Attacks for Tabular Data

Ixi-GEN: Efficient Industrial sLLMs through Domain Adaptive Continual Pretraining

Symbiosis: Multi-Adapter Inference and Fine-Tuning

Serving LLMs in HPC Clusters: A Comparative Study of Qualcomm Cloud AI 100 Ultra and NVIDIA Data Center GPUs

Learning Modular Exponentiation with Transformers

From High-SNR Radar Signal to ECG: A Transfer Learning Model with Cardio-Focusing Algorithm for Scenarios with Limited Data

ReDit: Reward Dithering for Improved LLM Policy Optimization

Flow based approach for Dynamic Temporal Causal models with non-Gaussian or Heteroscedastic Noises

Watermarking Autoregressive Image Generation

AssistedDS: Benchmarking How External Domain Knowledge Assists LLMs in Automated Data Science

Edit Flows: Flow Matching with Edit Operations

LeVo: High-Quality Song Generation with Multi-Preference Alignment

HauntAttack: When Attack Follows Reasoning as a Shadow

FuseUNet: A Multi-Scale Feature Fusion Method for U-like Networks

Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning

Leveraging Analytic Gradients in Provably Safe Reinforcement Learning

REOrdering Patches Improves Vision Models

Machine Unlearning under Overparameterization

Train with Perturbation, Infer after Merging: A Two-Stage Framework for Continual Learning

Balanced Token Pruning: Accelerating Vision Language Models Beyond Local Optimization

Autoencoding Random Forests

Towards Interpretability Without Sacrifice: Faithful Dense Layer Decomposition with Mixture of Decoders

LeCoDe: A Benchmark Dataset for Interactive Legal Consultation Dialogue Evaluation

Breaking mBad! Supervised Fine-tuning for Cross-Lingual Detoxification

Your Pre-trained LLM is Secretly an Unsupervised Confidence Calibrator

How Ensembles of Distilled Policies Improve Generalization in Reinforcement Learning

LLM-Explorer: A Plug-in Reinforcement Learning Policy Exploration Enhancement Driven by Large Language Models

RL Tango: Reinforcing Generator and Verifier Together for Language Reasoning

Text Generation Beyond Discrete Token Sampling

CLEVER: A Curated Benchmark for Formally Verified Code Generation

One-Step Offline Distillation of Diffusion-based Models via Koopman Modeling

CALM-PDE: Continuous and Adaptive Convolutions for Latent Space Modeling of Time-dependent PDEs

Shallow Flow Matching for Coarse-to-Fine Text-to-Speech Synthesis

Superposition Yields Robust Neural Scaling

Fair Clustering via Alignment

UMoE: Unifying Attention and FFN with Shared Experts

PRUNE: A Patching Based Repair Framework for Certifiable Unlearning of Neural Networks

Don't be lazy: CompleteP enables compute-efficient deep transformers

Fast-Slow Thinking GRPO for Large Vision-Language Model Reasoning

Integrating Structural and Semantic Signals in Text-Attributed Graphs with BiGTex

Token embeddings violate the manifold hypothesis

DIPLI: Deep Image Prior Lucky Imaging for Blind Astronomical Image Restoration

Deep Learning-Powered Electrical Brain Signals Analysis: Advancing Neurological Diagnostics

ExpertLens: Activation steering features are highly interpretable

Neural Attention Search

Toward Metaphor-Fluid Conversation Design for Voice User Interfaces

S$^2$-Diffusion: Generalizing from Instance-level to Category-level Skills in Robot Manipulation

DMWM: Dual-Mind World Model with Long-Term Imagination

BixBench: a Comprehensive Benchmark for LLM-based Agents in Computational Biology

Created by

Haebom

Author

Ludovico Mitchener, Jon M Laurent, Alex Andonian, Benjamin Tenmann, Siddharth Narayanan, Geemi P Wellawatte, Andrew White, Lorenzo Sani, Samuel G Rodriques

Outline

As large-scale language models (LLMs) and LLM-based agents demonstrate their potential to accelerate scientific research, the potential for autonomous AI-based discovery in bioinformatics is growing. This paper presents the Bioinformatics Benchmark (BixBench), a novel benchmark for measuring the biological data analysis capabilities of LLM-based agents. Comprised of over 50 real-world biological data analysis scenarios with nearly 300 open-ended questions, BixBench measures the ability of LLM-based agents to explore biological datasets, perform multi-step analysis trajectories, and interpret the analysis results. Using our own agent framework, we evaluated the performance of state-of-the-art LLMs, such as GPT-4o and Claude 3.5 Sonnet, achieving only 17% accuracy in open-answer settings and random-like performance in multiple-choice settings. BixBench aims to expose the current limitations of LLMs, spurring the development of agents capable of performing rigorous bioinformatics analysis and accelerating scientific discovery.