| { |
| "schema": "lifestack_baseline_eval_v1", |
| "note": "Base model eval (no LoRA). 50 episodes, same schedule as evaluate_and_plot. Per-episode rows omitted; re-run: python scripts/eval_baseline.py --output baseline_results.json", |
| "model": "Qwen/Qwen2.5-1.5B-Instruct", |
| "load_method": "transformers:Qwen/Qwen2.5-1.5B-Instruct", |
| "environment": "linux GPU server, HF fallback (Unsloth unavailable due to TRL import mismatch)", |
| "n_episodes": 50, |
| "mean_reward": -0.07, |
| "per_domain": { |
| "career": { "n": 7, "mean": -0.1429 }, |
| "finances": { "n": 7, "mean": 0.0 }, |
| "relationships": { "n": 6, "mean": 0.0 }, |
| "physical_health": { "n": 6, "mean": -0.1667 }, |
| "mental_wellbeing": { "n": 6, "mean": -0.25 }, |
| "time": { "n": 6, "mean": 0.0 }, |
| "transport_crisis": { "n": 6, "mean": 0.0 }, |
| "code_merge_crisis": { "n": 6, "mean": 0.0 } |
| }, |
| "all_domains_order": [ |
| "career", |
| "finances", |
| "relationships", |
| "physical_health", |
| "mental_wellbeing", |
| "time", |
| "transport_crisis", |
| "code_merge_crisis" |
| ] |
| } |
|
|
