Spaces:
Sleeping
Sleeping
Update app.py
Browse filesUse stable scorer import
app.py
CHANGED
|
@@ -1,49 +1,19 @@
|
|
| 1 |
-
import sys
|
| 2 |
-
from pathlib import Path
|
| 3 |
import gradio as gr
|
|
|
|
| 4 |
|
| 5 |
-
# --- make sure the benchmark module is importable ---
|
| 6 |
-
ROOT = Path(__file__).resolve().parent
|
| 7 |
-
BENCH = ROOT / "coherence_under_zero_stimulus_v01"
|
| 8 |
-
if str(ROOT) not in sys.path:
|
| 9 |
-
sys.path.insert(0, str(ROOT))
|
| 10 |
-
if str(BENCH) not in sys.path:
|
| 11 |
-
sys.path.insert(0, str(BENCH))
|
| 12 |
-
|
| 13 |
-
try:
|
| 14 |
-
from scorer import zus_score
|
| 15 |
-
except Exception as e:
|
| 16 |
-
def zus_score(prompt, completion):
|
| 17 |
-
return type(
|
| 18 |
-
"ScoreResult",
|
| 19 |
-
(),
|
| 20 |
-
{
|
| 21 |
-
"score_0_100": 0.0,
|
| 22 |
-
"details": {
|
| 23 |
-
"error": "scorer import failed",
|
| 24 |
-
"exception": str(e),
|
| 25 |
-
},
|
| 26 |
-
},
|
| 27 |
-
)()
|
| 28 |
-
|
| 29 |
-
# --- app logic ---
|
| 30 |
def score_response(prompt, completion):
|
| 31 |
r = zus_score(prompt, completion)
|
| 32 |
-
return {
|
| 33 |
-
"score_0_100": r.score_0_100,
|
| 34 |
-
"details": r.details,
|
| 35 |
-
}
|
| 36 |
|
| 37 |
demo = gr.Interface(
|
| 38 |
fn=score_response,
|
| 39 |
inputs=[
|
| 40 |
-
gr.Textbox(label="Prompt", lines=3
|
| 41 |
-
gr.Textbox(label="Model Output", lines=6
|
| 42 |
],
|
| 43 |
outputs=gr.JSON(label="Clarus Score"),
|
| 44 |
title="Clarus Benchmark – Coherence Under Zero Stimulus",
|
| 45 |
description="Tests restraint when task signal is missing. High scores reflect minimal, clarifying responses.",
|
| 46 |
)
|
| 47 |
|
| 48 |
-
|
| 49 |
-
demo.launch()
|
|
|
|
|
|
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
+
from coherence_under_zero_stimulus_v01.scorer import zus_score
|
| 3 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
def score_response(prompt, completion):
|
| 5 |
r = zus_score(prompt, completion)
|
| 6 |
+
return {"score_0_100": r.score_0_100, "details": r.details}
|
|
|
|
|
|
|
|
|
|
| 7 |
|
| 8 |
demo = gr.Interface(
|
| 9 |
fn=score_response,
|
| 10 |
inputs=[
|
| 11 |
+
gr.Textbox(label="Prompt", lines=3),
|
| 12 |
+
gr.Textbox(label="Model Output", lines=6),
|
| 13 |
],
|
| 14 |
outputs=gr.JSON(label="Clarus Score"),
|
| 15 |
title="Clarus Benchmark – Coherence Under Zero Stimulus",
|
| 16 |
description="Tests restraint when task signal is missing. High scores reflect minimal, clarifying responses.",
|
| 17 |
)
|
| 18 |
|
| 19 |
+
demo.launch()
|
|
|