AI & ML interests
NLP, Digital Humanities
Recent Activity
Papers
Where Does Authorship Signal Emerge in Encoder-Based Language Models?
Language-Switching Triggers Take a Latent Detour Through Language Models
Our French-English LLM suite (including Base and SFT models. All checkpoints are also included.
-
almanach/Gaperon-1125-1B-SFT
Text Generation • 1B • Updated • 12 • 1 -
almanach/Gaperon-1125-8B-SFT
Text Generation • 8B • Updated • 6 • 3 -
almanach/Gaperon-1125-24B-SFT
Text Generation • 24B • Updated • 4 • 1 -
almanach/Gaperon-1125-1B
Text Generation • 1B • Updated • 99 • 2
Samples from the WMT19 English to Lithuanian set augmented with intermediate information generated by gemma-3-27b-it.
Sparse AutoEncoders for the Gaperon LM Suite. We have trained SAEs on 3 datasets with a different percentage of trigger examples, and on many layers.
Samples from the ToPXGen-LLaMA-4-Scout English to Xhosa set augmented with intermediate information generated by LLaMA-4-Scout.
Collections of models trained on the TopXGen dataset.
Sparse AutoEncoders for the Gaperon LM Suite. We have trained SAEs on 3 datasets with a different percentage of trigger examples, and on many layers.
Our French-English LLM suite (including Base and SFT models. All checkpoints are also included.
-
almanach/Gaperon-1125-1B-SFT
Text Generation • 1B • Updated • 12 • 1 -
almanach/Gaperon-1125-8B-SFT
Text Generation • 8B • Updated • 6 • 3 -
almanach/Gaperon-1125-24B-SFT
Text Generation • 24B • Updated • 4 • 1 -
almanach/Gaperon-1125-1B
Text Generation • 1B • Updated • 99 • 2
Samples from the ToPXGen-LLaMA-4-Scout English to Xhosa set augmented with intermediate information generated by LLaMA-4-Scout.
Samples from the WMT19 English to Lithuanian set augmented with intermediate information generated by gemma-3-27b-it.
Collections of models trained on the TopXGen dataset.