INDEX

Explanations

references to harmful or damaging actions, particularly in a legal or critical context

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

destructive or disruptive

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

New Auto-Interp

Top Features by Cosine Similarity

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

Configuration

google/gemma-scope-9b-it-res/layer_9/width_131k/average_l0_121

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.9.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

tvguidetime

-0.61

exitRule

-0.59

Tikang

-0.59

⟬

-0.57

 estadounid

-0.55

DotNetBar

-0.54

 EventArgs

-0.49

 Мексичка

-0.49

 queſta

-0.48

vacy

-0.48

POSITIVE LOGITS

tructive

1.87

 destructive

1.86

destructive

1.79

 destru

1.02

destru

0.96

 disruptive

0.89

 destr

0.79

 constructive

0.76

 harmful

0.74

ruptive

0.73

Activations Density 0.005%

references to harmful or damaging actions, particularly in a legal or critical context

destructive or disruptive

No Comments

No Known Activations