影响力指数
论文质量、代表作、近期表现、广度与样本量置信度综合计算
92.82/100
前 0.4%
全站排名 #257
发表论文33 篇
平均评分
年均产出11.0 篇/年
Razvan Pascanu
研究方向
Continual Learning/Lifelong learning · Graph Nets · Deep Reinforcement Learning · Optimization · Deep Learning · RNN
13
From Markov to Laplace: How Mamba In-Context Learns Markov Chains
ICLR 2026Oral
16
MesaNet: Sequence Modeling by Locally Optimal Test-Time Training
ICLR 2026Poster
18
Lattice: Learning to Efficiently Compress the Memory
ICLR 2026Rejected
二作18
LLMs are Greedy Agents: Effects of RL Fine-tuning on Decision-Making Abilities
ICLR 2026Poster
通讯9
Fine-Tuned In-Context Learners
ICLR 2026Rejected
三作4
Unpacking Softmax: How Logits Norm Drives Representation Collapse, Compression and Generalization
ICLR 2026Withdrawn
通讯25
Attention as a Hypernetwork
ICLR 2025Oral
通讯18
Meta-learning how to Share Credit among Macro-Actions
NeurIPS 2025Poster
三作23
Plasticity as the Mirror of Empowerment
NeurIPS 2025Spotlight
15
How do language models learn facts? Dynamics, curricula and hallucinations
COLM 2025Poster
10
MS-SSM: A Multi-Scale State Space Model for Efficient Sequence Modeling
COLM 2025Poster
16
Why do LLMs attend to the first token?
COLM 2025Poster
32
softmax is not enough (for sharp out-of-distribution)
ICLR 2025Rejected
通讯31
Round and Round We Go! What makes Rotary Positional Encodings useful?
ICLR 2025Poster
13
Softmax is not Enough (for Sharp Size Generalisation)
ICML 2025Poster
通讯11
A Large Recurrent Action Model: xLSTM enables Fast Inference for Robotics Tasks
ICML 2025Poster
26
RAT: Bridging RNN Efficiency and Attention Accuracy via Chunk-based Sequence Modeling
NeurIPS 2025Poster
三作22
Is multitask learning all you need in continual learning?
ICLR 2025Rejected
通讯21
A Large Recurrent Action Model: xLSTM enables Fast Inference for Robotics Tasks
ICLR 2025Rejected
18
Transformers meet Neural Algorithmic Reasoners
ICLR 2025Rejected
12
Retrieval-Augmented Decision Transformer: External Memory for In-context RL
ICLR 2025Rejected
18
Torque-Aware Momentum
ICLR 2025Rejected