Spaces:

st192011
/

Entropy-Perplexity-Routing

Running

st192011 commited on 10 days ago

Commit

e793199

verified ·

1 Parent(s): c3f8ec8

Update app.py

Files changed (1) hide show

app.py CHANGED Viewed

@@ -231,7 +231,7 @@ with gr.Blocks(theme=gr.themes.Base(), css=simplified_css) as demo:
 ---
 ### 1. Introduction & Experimental Setup
-The objective of this study was to evaluate and optimize the zero-shot reasoning capabilities of a Small Language Model (SLM) on multiple-choice question answering.
 * **Dataset:** The CAIS/MMLU (Massive Multitask Language Understanding) benchmark, specifically utilizing randomized validation splits across diverse academic disciplines.
 * **Methodology:** We compared traditional heuristic prompt engineering methods against a dynamic, model-agnostic routing framework that switches between standard token generation and sequence likelihood evaluation (Perplexity).

 ---
 ### 1. Introduction & Experimental Setup
+The objective of this study was to evaluate and optimize the zero-shot reasoning capabilities of a Small Language Model (google/gemma-4-E2B) on multiple-choice question answering.
 * **Dataset:** The CAIS/MMLU (Massive Multitask Language Understanding) benchmark, specifically utilizing randomized validation splits across diverse academic disciplines.
 * **Methodology:** We compared traditional heuristic prompt engineering methods against a dynamic, model-agnostic routing framework that switches between standard token generation and sequence likelihood evaluation (Perplexity).