UniClawBench: A Universal Benchmark for Proactive Agents on Real-World Tasks

▲ 25 🎓 HKU MMLab

• UniClawBench introduces a capability-driven benchmark for evaluating proactive agents in real-world environments using live Docker container evaluatio

Ideas Have Genomes: Benchmarking Scientific Lineage Reasoning and Lineage-Grounded Idea Ge

▲ 25 🎓 Shanghai Jiao Tong University

• A benchmark for scientific lineage reasoning and idea generation is introduced, organizing scientific works as genetic-like Idea Genome objects and ev

LongE2V: Long-Horizon Event-based Video Reconstruction, Prediction, and Frame Interpolatio

▲ 22 🎓 National Yang Ming Chiao Tung University

• LongE2V enables high-quality video recovery from sparse event streams by leveraging pre-trained video diffusion priors and addressing temporal stabili

DrugGen 2: A disease-aware language model for enhancing drug discovery

▲ 14 🎓 Isfahan University of Medical Sciences

• DrugGen-2 generates small molecules conditioned on disease ontology and target protein sequences through fine-tuning GPT-2 with supervised learning an

Enhancing In-context Panoramic Generation via Geometric-aware Pretraining

▲ 14 🏢 Insta360 Research

• Canvas360 is a two-stage framework for in-context panoramic generation that combines geometry-aware pretraining with fine-tuning, featuring a large-sc

OpenCoF: Learning to Reason Through Video Generation

▲ 9 🏢 ByteDance

• OpenCoF framework introduces a reasoning video dataset and model that improve temporal reasoning through diverse supervision and explicit reasoning to

OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Genera

▲ 6 🏢 Meituan

• OPSD-V enhances few-step autoregressive video diffusion models by using real long-video data for temporal context during training, providing dense tra

Remember When It Matters: Proactive Memory Agent for Long-Horizon Agents

▲ 5 🏢 Meta AI

• In long-horizon tasks, decision-relevant state is often scattered across an expanding trajectory, while the action agent must surface it and act.
• As trajectories grow, task requirements, environment facts, prior attempts, diagnoses, and open subgoals can be buried in the context window or pushed
• We call this failure mode "behavioral state decay".
• We study memory as an active intervention mechanism rather than passive retrieval.

Closed-form fractional radial links for elliptical Mahalanobis discriminant analysis

🏛️ Cherkasy State Business College AI & Machine Learning

• The need for interpretable machine learning models is increasingly critical as their applications expand across various
• A method for steering neural network training through interpretable constraints based on partial dependence.

Validity of LLMs as data annotators: AMALIA on authority

🏛️ Universidade Lusófona AI & Machine Learning

• Large language models (LLMs) often struggle with reasoning and logical consistency, leading to unreliable outputs.
• A framework for quantifying uncertainty, coherence, and robustness of LLMs using a graph-based approach.

Steering Neural Network Training through Interpretable Constraints Based on Partial Depend

🏛️ University of Liège AI & Machine Learning

• The efficiency of model evaluation is hindered by fixed-size benchmarks that do not adapt to diverse evaluation objectiv
• A method for efficient model evaluation that determines the optimal amount of data needed for testing.

GradInf: Gradient Estimation as Probabilistic Inference

🏛️ Carnegie Mellon University Software & Programming

• Gradient estimation in probabilistic programs is notoriously difficult and has diverse applications in scientific comput
• A novel method for gradient estimation framed as probabilistic inference.

Locality of Curve-Decoding and Improved Proximity Gaps

🏛️ Massachusetts Institute of Technology Theory & Algorithms

• Understanding the locality of curve-decoding in error-correcting codes is crucial for improving decoding efficiency.
• A study on proximity gaps in the context of Interactive Oracle Proofs (IOPs) and Succinct Non-interactive Arguments of Z

Sculptable Mesh Structures for Room-Scale Form-Finding

🏛️ Carnegie Mellon University Applications

• Designing physical structures effectively within the constraints of a computer interface is challenging.
• A sculptable mesh structure approach that facilitates room-scale form-finding through interactive design.

MMGenre: Benchmarking Singing Voice Synthesis across Multiple Musical Genres

🏛️ Renmin University of China Other CS

• Singing voice synthesis (SVS) has limited generalization across diverse musical genres, affecting its practical applicab
• A benchmarking framework, MMGenre, designed to evaluate SVS performance across multiple musical genres.

Estimating the Stochastic Discount Factor from Option Prices and Predicting the Equity Pre

🏛️ The University of Tokyo Quantitative Finance

• Estimating the stochastic discount factor (SDF) from option prices is complex and often inaccurate.
• A new framework that scales the SDF by time-varying volatility using market data from S&P 500 options.

Estimating Causal Effects from Data Generated by Stochastic Algorithms

🏛️ Stanford University Economics

• Estimating causal effects from data generated by stochastic algorithms presents unique challenges.
• The paper proposes a method for estimating causal effects in contexts where content is selected based on user characteri