INDEX

Explanations

statements where someone agrees with a particular viewpoint or opinion

oai_token-act-pair · gpt-3.5-turbo

expressions of agreement

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 procedural

-0.65

 hide

-0.59

 idle

-0.58

 imper

-0.57

ãĤ¹

-0.57

ser

-0.56

 hidden

-0.56

 camouflage

-0.56

 Gener

-0.55

 tactical

-0.55

POSITIVE LOGITS

 agrees

3.60

 disagrees

2.25

 concedes

1.71

 approves

1.70

 acknowledges

1.69

 confirms

1.69

 agreed

1.66

 accepts

1.63

 agree

1.60

 admits

1.58

Activations Density 0.018%

statements where someone agrees with a particular viewpoint or opinion

expressions of agreement

No Comments

No Known Activations