improve: abstention penalty, better prompt, mixed curriculum, more steps

evaluate: switch to 0.5B model comparison, 200 episodes

fix: set N_EPISODES=200 constant (was still 30)

update: increase evaluation to 200 episodes per model

fix: parse_action confidence bug, numeric answers bug, missing reasoning field bug

fix: debug model output parsing in evaluation

add: evaluate 1.5B base vs trained, upload chart to HF Hub

add: evaluate 1.5B base vs trained, upload charts

update: compare 0.5B vs 1.5B trained models

update: evaluate retrained model, upload charts to HF Hub

update: evaluate on 200 episodes each for more reliable results

fix: remove erroneous del model outside evaluate_model scope

fix: free GPU memory between model evaluations

add: evaluation script - base vs trained model comparison