INDEX

Explanations

instructions or guidelines that emphasize consistency and frequency in actions or behaviors

oai_token-act-pair · gpt-4o-mini Triggered by @bot

always / never

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Said

-0.45

 propOrder

-0.43

 Soon

-0.43

 Said

-0.41

läng

-0.41

 maybe

-0.40

NR

-0.40

Initial

-0.40

Various

-0.39

 impati

-0.39

POSITIVE LOGITS

 ALWAYS

0.81

always

0.81

 always

0.79

ALWAYS

0.74

 Always

0.73

Always

0.71

deauna

0.69

 sempre

0.68

 zawsze

0.68

 alltid

0.67

Activations Density 0.026%

instructions or guidelines that emphasize consistency and frequency in actions or behaviors

always / never

No Comments

No Known Activations