INDEX

Explanations

violent actions and physical harm

oai_token-act-pair · gpt-3.5-turbo

references to acts of violence or severe harm

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

entric

-0.71

uci

-0.69

 Collabor

-0.65

issions

-0.63

ty

-0.61

TP

-0.60

 extension

-0.60

impl

-0.60

 Mutual

-0.59

AE

-0.59

POSITIVE LOGITS

 beaten

3.79

 beat

1.93

 beating

1.85

beat

1.59

 battered

1.55

 slain

1.44

 defeated

1.42

Beat

1.39

 beats

1.34

 bruised

1.33

Activations Density 0.018%

violent actions and physical harm

references to acts of violence or severe harm

No Comments

No Known Activations