TIDE Analysis - Live Results Dashboard

🔬 Research Status: ACTIVE

Data collection started: 2025-07-22T14:57:32.317876

Latest update: 2025-07-22 15:53:42

This dashboard updates as new experimental sessions are completed

📊 Current Results by Model

Note on Coherence Scores: These scores measure how consistent each model is with itself when responding to similar prompts. Higher scores indicate more predictable, consistent responses; lower scores indicate more varied, creative responses.

gemini-1.5-flash

71.5% self-consistency

Sessions

210

Responses

Feature Activation Preference (based on actual content):

Internal
11%

External
73%

Concrete
16%

Pattern Distribution:

AAFC: 191 (91%) CCFC: 11 (5%) AAFS: 8 (4%)

Shows 3 distinct pattern types

claude-3-haiku-20240307

55.1% self-consistency

Sessions

210

Responses

Feature Activation Preference (based on actual content):

Internal
28%

External
54%

Concrete
18%

Pattern Distribution:

AAFC: 122 (58%) CCFC: 61 (29%) AAFS: 14 (7%) CCFS: 13 (6%)

Shows 4 distinct pattern types

gpt-3.5-turbo

38.3% self-consistency

Sessions

210

Responses

Feature Activation Preference (based on actual content):

Internal
32%

External
39%

Concrete
29%

Pattern Distribution:

AAFC: 121 (58%) CCFC: 41 (20%) AAFS: 41 (20%) CCFS: 5 (2%) AADC: 2 (1%)

Shows 5 distinct pattern types (most diverse)

⚡ Understanding These Results

Coherence Score: Measures internal consistency - how similarly a model responds to similar prompts. Not a measure of quality or correctness.

Dimensional Preferences: Based on actual feature activation in responses, showing which semantic dimensions each model naturally emphasizes.

Pattern Types: Different response signatures identified. More pattern diversity may indicate more flexible processing.

Current n=21 sessions. Results will become more robust with additional data collection.

TIDE Analysis: Live Results Dashboard

🔬 Research Status: ACTIVE

📊 Current Results by Model

gemini-1.5-flash

Feature Activation Preference (based on actual content):

Pattern Distribution:

claude-3-haiku-20240307

Feature Activation Preference (based on actual content):

Pattern Distribution:

gpt-3.5-turbo

Feature Activation Preference (based on actual content):

Pattern Distribution:

⚡ Understanding These Results