Back to corpus
experimentexperiment writeup candidatescore 18

5. Results and Analysis

**Training Loss Evolution**: - Initial loss: 1449.73 - Convergent loss: ~1418.69 (validation) - Convergence rate: Exponential with λ ≈ 0.023 - Stability: No oscillations or divergence

Full HTML reader

Read the full artifact

Open in new tab

Extracted abstract or opening context

**Training Loss Evolution**: - Initial loss: 1449.73 - Convergent loss: ~1418.69 (validation) - Convergence rate: Exponential with λ ≈ 0.023 - Stability: No oscillations or divergence **Conservation Constraint Satisfaction**: - Measure preservation: 0.87 ± 0.03 - Ergodic stability: 0.91 ± 0.02 - Information conservation: 0.84 ± 0.04 - Hamiltonian conservation: 0.89 ± 0.03 **Overall Coordinate Accuracy**: - Root Mean Square Error: 0.445 - Mean Absolute Error: 0.312 - Coordinate prediction confidence: 0.889 **Pattern Consistency Metrics**: - Intra-individual consistency: 0.823 - Inter-individual distinctiveness: 0.756 - Pattern stability over time: 0.891 - Response predictability: 0.734 **Attention Allocation Learning**: - Attention weight consistency: 0.867 - Context utilization accuracy: 0.798 - Focus pattern recognition: 0.812

Promotion decision

What has to happen next

Attach run IDs, datasets, metrics, and reproduction commands.

Why this is not always a full paper yet

Corpus pages are public-safe readers for discovered workspace artifacts. They are not automatically final papers. A corpus item becomes a polished paper only after the editable source, evidence checkpoints, references, figures, render path, and release status are attached through the paper schema.