INDEX

Explanations

mentions of events or actions that provoke strong reactions and protests

oai_token-act-pair · gpt-3.5-turbo

instances of events or actions that lead to public reactions or protests

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 Poss

-0.71

pure

-0.64

 assimil

-0.64

 faithfully

-0.63

soever

-0.62

 isEnabled

-0.62

 Proper

-0.60

enery

-0.59

 Present

-0.59

resents

-0.59

POSITIVE LOGITS

 rethink

0.96

 backlash

0.84

ire

0.84

 mockery

0.76

lash

0.75

 closure

0.73

 reconsider

0.73

 resignation

0.72

 lockdown

0.72

 inquiry

0.71

Activations Density 0.539%

mentions of events or actions that provoke strong reactions and protests

instances of events or actions that lead to public reactions or protests

No Comments

No Known Activations