INDEX

Explanations

elements related to surprises or unexpected events

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

surprise

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

 recueil

-0.48

 reposer

-0.45

 drapeau

-0.44

 abstrato

-0.41

 ujednoznacz

-0.41

 Occidente

-0.39

remacy

-0.37

 progrès

-0.37

 plainte

-0.37

Descrizione

-0.36

POSITIVE LOGITS

 surprise

0.81

parsedMessage

0.63

surprise

0.63

 Surprise

0.60

 snippetHide

0.59

Surprise

0.58

 surprises

0.57

thâu

0.55

 bezeichneter

0.54

 surprised

0.53

Activations Density 0.027%

elements related to surprises or unexpected events

surprise

No Comments

No Known Activations