INDEX

Explanations

the structure of news articles

oai_token-act-pair · gpt-3.5-turbo

sequences of underscores or similar symbols, denoting sections or breaks in the text

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

 Dynamics

-0.76

 Instr

-0.72

ovan

-0.71

 Myster

-0.70

 Noir

-0.70

 Diver

-0.69

oms

-0.67

ways

-0.65

 Exile

-0.65

 Vine

-0.64

POSITIVE LOGITS

vu

0.89

___

0.82

SOURCE

0.81

DOWN

0.77

HAEL

0.76

PLE

0.75

enhagen

0.74

kw

0.71

POS

0.71

quote

0.68

Activations Density 0.017%

the structure of news articles

sequences of underscores or similar symbols, denoting sections or breaks in the text

No Comments

No Known Activations