INDEX

Explanations

instances of suspensions

oai_token-act-pair · gpt-3.5-turbo

instances of the word "suspended" in contexts related to penalties or disciplinary actions

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

rics

-0.79

åĮ

-0.77

 Generations

-0.71

bda

-0.69

 originals

-0.69

swick

-0.67

MAP

-0.67

raltar

-0.67

ebus

-0.66

PRES

-0.66

POSITIVE LOGITS

 suspending

1.09

 suspensions

1.07

 suspended

1.05

 suspension

1.02

 indefinitely

1.00

 disbelief

0.98

 suspend

0.93

 susp

0.84

ishment

0.81

owship

0.80

Activations Density 0.023%

instances of suspensions

instances of the word "suspended" in contexts related to penalties or disciplinary actions

No Comments

No Known Activations