INDEX

Explanations

words related to disciplinary actions, reprimands, and investigations

oai_token-act-pair · gpt-3.5-turbo

terms related to disciplinary actions or reprimands

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

population

-0.78

pop

-0.74

 originals

-0.73

ãĤª

-0.69

soDeliveryDate

-0.68

sonian

-0.66

ortment

-0.65

mix

-0.64

ITNESS

-0.63

ãĤ¤ãĥĪ

-0.63

POSITIVE LOGITS

 disciplinary

0.99

 reprim

0.97

 levied

0.86

 leveled

0.81

report

0.79

 rebuke

0.74

HAEL

0.73

 Inquiry

0.73

 hearings

0.71

alties

0.71

Activations Density 0.051%

words related to disciplinary actions, reprimands, and investigations

terms related to disciplinary actions or reprimands

No Comments

No Known Activations