Results: SIBA Meristem against its base model

README.md

@@ -75,9 +75,18 @@ scores against this baseline, not against 25 %.
 
 ## Results
 
-The leaderboard is being filled in and is not complete yet. Runs so far only cover part
-of `belebele-az`, so no scores are published here until every config has been run for
-every model. The per-item results live in the GitHub repository.
+Only models that answered every question are listed. Both are 3B models at Q4_K_M
+through Ollama, zero-shot, temperature 0.
+
+| model | belebele-az | include-az | tumlu-az |
+|---|---:|---:|---:|
+| Qwen2.5-3B-Instruct | 45.2 % | 36.9 % | 36.0 % |
+| [SIBA Meristem 1.0](https://huggingface.co/aghasalim/siba-meristem-1.0) (Qwen2.5-3B fine-tuned on Azerbaijani) | 38.9 % | 40.0 % | 37.7 % |
+| always the most common answer | 27.9 % | 26.3 % | 42.7 % |
+
+The fine-tune gains on exam questions and loses 6.3 points on reading comprehension.
+On `tumlu-az` neither model beats always answering D. Partial runs of larger models
+and the raw counts are in the [GitHub repository](https://github.com/aghasalim/azbench).
 
 ## Sources