33 papers
KVpop -- Key-Value Cache Compression with Predictive Online Pruning
Lukas Hauzenberger, Niklas Schmidinger, Anamaria-Roberta Hartl +5
Key-value (KV) cache growth is a major bottleneck in autoregressive decoding, as memory and bandwidth scale linearly with context length. Existing KV eviction methods often rely on…
Effective Distillation to Hybrid xLSTM Architectures
Lukas Hauzenberger, Niklas Schmidinger, Thomas Schmied +7
There have been numerous attempts to distill quadratic attention-based large language models (LLMs) into sub-quadratic linearized architectures. However, despite extensive research…
TiRex-2: Generalizing TiRex to Multivariate Data and Streaming
Patrick Podest, Marco Pichler, Elias Bürger +7
We introduce TiRex-2, a recurrent xLSTM-based time series foundation model that generalizes the univariate TiRex to multivariate forecasting with both past and future covariates. R…
On Subquadratic Architectures: From Applications to Principles
Anamaria-Roberta Hartl, Levente Zólyomi, David Stap +6
Transformers dominate modern sequence modeling, but their quadratic attention incurs substantial computational cost. Subquadratic architectures offer a scalable alternative. Howeve…
RREDCoT: Segment-Level Reward Redistribution for Reasoning Models
Mykyta Ielanskyi, Kajetan Schweighofer, Lukas Aichberger +1
Recent advancements in reasoning language models have been driven by Reinforcement Learning (RL) fine-tuning. Most often, these rely on the Group Relative Policy Optimization (GRPO…
Unlocking the Working Memory of Large Language Models for Latent Reasoning
Lukas Aichberger, Sepp Hochreiter
To improve the reasoning capabilities of large language models, test-time compute is typically scaled by generating intermediate tokens before the final answer. However, this coupl…