4 papers
When Attention Goes Blind: Numerical Failure in ALiBi Positional Encodings
Christopher Schröder, Christopher Schröder, Lukas Gienapp +3
We identify a previously overlooked failure mode of ALiBi positional encoding: its linear bias scaling underflows floating-point precision, which zeroes out a large fraction of att…
Topic-Specific Classifiers are Better Relevance Judges than Prompted LLMs
Lukas Gienapp, Martin Potthast, Andrew Yates +2
The unjudged document problem, where systems that did not contribute to the original judgement pool may retrieve documents without a relevance judgement, is a key obstacle to the r…
The German Commons - 154 Billion Tokens of Openly Licensed Text for German Language Models
Lukas Gienapp, Christopher Schröder, Stefan Schweter +5
Large language model development relies on large-scale training corpora, yet most contain data of unclear licensing status, limiting the development of truly open models. This prob…
The Viability of Crowdsourcing for RAG Evaluation
Lukas Gienapp, Tim Hagen, Maik Fröbe +4
How good are humans at writing and judging responses in retrieval-augmented generation (RAG) scenarios? To answer this question, we investigate the efficacy of crowdsourcing for RA…