INDEX

Explanations

references to spokespeople or representatives in media contexts

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

official spokesperson

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

 Sault

-0.44

 Yarmouth

-0.44

 Duluth

-0.43

 BoxFit

-0.43

 snippetHide

-0.43

cstdlib

-0.43

detroit

-0.42

 ruff

-0.42

 Highland

-0.41

jeeling

-0.41

POSITIVE LOGITS

 spokesman

1.02

 spokesperson

0.94

 Sprecher

0.85

 spokeswoman

0.84

 portavoz

0.82

 spokes

0.79

 spokespersons

0.77

 Spokes

0.73

voces

0.65

 speaker

0.59

Activations Density 0.003%

references to spokespeople or representatives in media contexts

official spokesperson

No Comments

No Known Activations