EvalExplorer classifier
📊
Small models that replace a big-LLM classifier
EvalExplorer document classifier
Small models that classify an evaluation report's first pages into approach, type, temporality, themes and countries. Data, adapters, run reports.
Note Leaderboards for the Baobab Tech fine-tuning experiments
baobabtech/evalexplorer-classify-experiments
Viewer • Updated • 7.43k • 611Note One report per run, plus the leaderboard
baobabtech/evalexplorer-data
Viewer • Updated • 225k • 73Note Source export of the EvalExplorer database
baobabtech/evalexplorer-classify-qwen3.5-4b-sft
Updated • 3Note SFT - mean field score 0.847
baobabtech/evalexplorer-classify-gemma-4-26b-a4b-sft
Updated • 8Note SFT - mean field score 0.844
baobabtech/evalexplorer-classify-qwen3.5-2b-sft
Updated • 6Note SFT - mean field score 0.842
baobabtech/evalexplorer-classify-gemma-4-e4b-sft
Updated • 1Note SFT - mean field score 0.830
baobabtech/evalexplorer-classify-gemma-4-e2b-sft
Updated • 14Note SFT - mean field score 0.815
baobabtech/evalexplorer-classify-lfm2.5-1.2b-sft
Updated • 1Note SFT - mean field score 0.802
baobabtech/evalexplorer-classify-lfm2.5-350m-sft
Updated • 3Note SFT - mean field score 0.792
baobabtech/evalexplorer-classify-gliner2.5-base
0.2B • Updated • 1Note GLINER-FINETUNE - mean field score 0.578
baobabtech/evalexplorer-classify-gliner2.5-small
73.9M • Updated • 3Note GLINER-FINETUNE - mean field score 0.573
baobabtech/evalexplorer-classify-adapters
UpdatedNote Experimental adapters (GRPO variants), one subfolder each
baobabtech/evalexplorer-classify-gguf
25B • Updated • 158Note GGUF exports for llama.cpp: Q8_0, Q5_K_M, Q4_K_M per adapter