Cognitive Nutrition — Evaluation Explorer
Alexandria Aeternum 10K | Llama 3.2 11B Vision-Instruct | LoRA r=32 | 100 eval images x 3 models x 4 tracks
Experiment: Feb 12-16, 2026 | ~94 hours on a single NVIDIA L4 | ~$200 compute
Overview
Training
VULCA (L1-L5)
POPE
Response Quality
CogBench
Counterfactual
Semantic Density
←
→
Navigate |
1
-
8
Tracks