Adarsh290406/gov-scheme-env

fix: run both uvicorn server and inference.py via start.sh

fix: run both uvicorn server and inference.py via start.sh

fix: run both uvicorn server and inference.py via start.sh

fix: run both uvicorn server and inference.py via start.sh

fix: run both uvicorn server and inference.py via start.sh

Inject sys.stderr debug dump to trace platform grader constraints

Final Fix: Forced 2dp float clamping for all logs & patched missing fatal keys

Final: Universal 0.05-0.95 clamping in API and Inference logs

Final Fix: Forced 2dp float clamping (0.05-0.95) for all logs

Hard-clamp task scores and eliminate 0.0/1.0 footprints

Hard-clamp task scores and eliminate 0.0/1.0 footprints

Fix: Strict (0.05, 0.95) clamping to pass Task Validation range check

Fix: Strict (0.05, 0.95) clamping to pass Task Validation range check

Fix: Strict (0.05, 0.95) clamping to pass Task Validation range check

Fix: Strict (0.05, 0.95) clamping to pass Task Validation range check

Rewrite internal grading schemas to strictly map natively within limits

Rewrite internal grading schemas to strictly map natively within limits

Rewrite internal grading schemas to strictly map natively within limits

Rewrite internal grading schemas to strictly map natively within limits

Rewrite internal grading schemas to strictly map natively within limits

Rewrite internal grading schemas to strictly map natively within limits

Fix score clamping boundary conditions

Hard clamp all Step and End reward bounds for strictly mathematically open validation

Fix score boundaries for strict RL task validation, enable LLM proxy test

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

inference.py

add TASK markers

final submission clean version

final submission clean version