INDEX

Explanations

words related to disciplinary actions such as suspension

oai_token-act-pair · gpt-3.5-turbo

instances of suspension or related disciplinary actions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

PT

-0.89

eds

-0.77

atche

-0.76

ipped

-0.75

rics

-0.73

PRES

-0.73

Drive

-0.73

atered

-0.73

rouse

-0.71

xml

-0.71

POSITIVE LOGITS

 suspended

1.43

 suspending

1.39

 suspend

1.17

 suspensions

1.14

 suspension

1.11

 susp

0.93

 lockdown

0.90

 indefinitely

0.88

 gobl

0.88

©¶æ¥µ

0.87

Activations Density 0.011%

words related to disciplinary actions such as suspension

instances of suspension or related disciplinary actions

No Comments

No Known Activations