AI Research Brief
Search
Methodology
中文
Prism Trains 2.5× Faster and Improves Quality
15 selected from 356 papers
Featured
Prism: Dynamic Sparse Attention for Native 2K Joint Video-Audio Generation Model Training
score 6
入选 HF Daily Papers; HF 热度: 4 upvotes (+1); 有代码实现
Also Worth Noting
Invisible Ink, Visible Lies: How Production Watermarking Causes LLMs to Hallucinate
score 4
关键词(2): production, retrieval-augmented; 顶会接收: NeurIPS
EDISCO: Equivariant DIScrete Diffusion for Euclidean Combinatorial Optimization
score 4
关键词(2): edge, synthetic data; 顶会接收: NeurIPS
Triggering Generalist Reasoning via Predictive Uncertainty for Dual-System VLA
score 4
关键词(3): real-time, reasoning, vision-language; 顶会接收: NeurIPS
LogSig-SSM: Time-Series Modelling with Multi-Scale Log-Signature Compression for State-Space Models
score 4
关键词(2): compression, mamba; 顶会接收: NeurIPS
When and What to Prune? Stage-Aware Visual Token Pruning for Efficient VLA
score 4
关键词(2): pruning, vision-language; 顶会接收: NeurIPS
ASCENT: Online Test-Time Training of Long-Horizon Agents via Self-Distillation of Verified Experience
score 7
入选 HF Daily Papers; HF 热度: 13 upvotes (+3); 关键词(4): distillation, deployment, agentic, reasoning
GNN-CB: A Graph Neural Network Competition Benchmark for Human and LLM Evaluation
score 4
关键词(3): edge, coding, reasoning; 顶会接收: EMNLP
Expanding LLM Reasoning
score 4
关键词(1): reasoning; 顶会接收: NeurIPS
ColdDDI: Evaluating Knowledge Utilization in Cold-Start Drug-Drug Interaction Prediction
score 4
关键词(1): edge; 顶会接收: NeurIPS
Usage-Modulated Sentiment Representations in Large Language Models
score 3
顶会接收: EMNLP
Tracing a Sparse Emotion-Control Circuit in LLM-Based Text-to-Speech
score 3
顶会接收: EMNLP
Order Matters: Competition-Guided Query Ordering for RNN-Based Object Detection
score 3
顶会接收: NeurIPS
TeleTune: Evolving Agent Skills From Offline Telemetry
score 3
机构: Microsoft
What Does an Observability Foundation Model Know?
score 3
顶会接收: NeurIPS