INDEX

Explanations

references to different levels or parts of a structure, with a particular focus on the upper part

oai_token-act-pair · gpt-3.5-turbo

references to the concept of "upper" related to classes, bodies, or positions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

tnc

-0.93

 Machina

-0.89

printf

-0.75

uality

-0.74

eln

-0.71

 Killer

-0.69

 Frankenstein

-0.69

 Hurricanes

-0.67

Pok

-0.66

HCR

-0.66

POSITIVE LOGITS

most

1.02

 shaft

0.87

bridge

0.81

jaw

0.79

 earners

0.78

 thigh

0.77

 precedence

0.76

ĸļå£«

0.76

 bounds

0.75

 respiratory

0.74

Activations Density 0.014%

references to different levels or parts of a structure, with a particular focus on the upper part

references to the concept of "upper" related to classes, bodies, or positions

No Comments

No Known Activations