INDEX

Explanations

phrases related to defending or saving off negative outcomes or threats

oai_token-act-pair · gpt-3.5-turbo

terms related to actions that mitigate or prevent negative outcomes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

rian

-0.61

ynski

-0.58

igma

-0.57

å¥³

-0.57

Scot

-0.56

 Supplemental

-0.55

HUD

-0.55

bsite

-0.55

agents

-0.55

 Brewer

-0.54

POSITIVE LOGITS

off

1.91

off

1.48

offs

1.44

Off

1.35

Off

1.31

OFF

1.27

OFF

1.16

 dividends

0.77

 goodbye

0.76

 shoulders

0.71

Activations Density 0.301%

phrases related to defending or saving off negative outcomes or threats

terms related to actions that mitigate or prevent negative outcomes

No Comments

No Known Activations