5 citations · 13 across the 13 of their papers we have counts for
6 papers · 1 filter
Kolibri: A Sovereign European Model on the Pareto Frontier
Aleph Alpha, :, Moritz Althaus +65
This report introduces Kolibri, an English-German mixture-of-experts transformer with 78.1B total parameters and 3.46B active parameters per token, released as open weights under t…
Exact Quantile Balancing and Load-Error Injection for Mixture-of-Experts
Pit Neitemeier, Jiaze Li, Alessio Serra +2
Mixture-of-Experts (MoE) training requires global load balance to prevent expert under-utilization and local balance for efficient expert-parallel execution. Existing distributed Q…
Good Pretraining, Bad SFT: Checkpoint Quality Across the Training Stack
Sohir Maskey, Philipp Scholl, Jonas Knupp +2
Language-model checkpoints are commonly selected by pretraining loss or benchmark scores, assuming that the highest-scoring checkpoint will remain the best starting point for subse…
Studying quantization trade-offs for efficient inference deployment in machine translation
Jim Zhao, Sohir Maskey, Koen Oostermeijer +2
Deploying large language models in realistic server environments poses challenges, as the system needs to provide high-quality responses with low latency. Quantization is a common…
The Price of Robustness: Stable Classifiers Need Overparameterization
Jonas von Berg, Adalbert Fono, Massimiliano Datres +2
The relationship between overparameterization, stability, and generalization remains incompletely understood in the setting of discontinuous classifiers. We address this gap by est…
1-Bit Wonder: Improving QAT Performance in the Low-Bit Regime through K-Means Quantization
Sohir Maskey, Constantin Eichenberg, Johannes Messner +1
Quantization-aware training (QAT) is an effective method to drastically reduce the memory footprint of LLMs while keeping performance degradation at an acceptable level. However, t…