INDEX

Explanations

phrases related to control, authority, and conflict

oai_token-act-pair · gpt-3.5-turbo

conjunctions and linking terms that connect ideas and clauses in the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 appro

-0.73

ĸ

-0.64

iren

-0.59

emn

-0.58

 surreal

-0.58

©¶æ

-0.58

 endorsement

-0.57

Aid

-0.56

 flagship

-0.56

amic

-0.56

POSITIVE LOGITS

 selves

0.75

vous

0.72

 Templ

0.71

 Bullets

0.67

azy

0.66

itsch

0.63

 mates

0.62

arent

0.60

inki

0.59

iblings

0.59

Activations Density 0.574%

phrases related to control, authority, and conflict

conjunctions and linking terms that connect ideas and clauses in the text

No Comments

No Known Activations