INDEX

Explanations

phrases indicating reasoning or logical argumentation

oai_token-act-pair · gpt-3.5-turbo

phrases that refer to different types or "lines" of reasoning or argumentation

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

»Ĵ

-0.68

Reviewer

-0.66

gyn

-0.63

 Pradesh

-0.62

huge

-0.61

 Powers

-0.61

 avail

-0.59

bear

-0.58

hunt

-0.57

orporated

-0.56

POSITIVE LOGITS

 scrimmage

0.90

 sight

0.72

 succession

0.72

 questioning

0.70

 descent

0.70

erie

0.68

otomy

0.65

backer

0.64

 reasoning

0.63

ographs

0.62

Activations Density 0.088%

phrases indicating reasoning or logical argumentation

phrases that refer to different types or "lines" of reasoning or argumentation

No Comments

No Known Activations