INDEX

Explanations

things that are more than a specified limit or standard

oai_token-act-pair · gpt-3.5-turbo

instances of the word "exceed" and its variations, indicating thresholds or limits being surpassed

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

Sham

-0.79

Ale

-0.73

pring

-0.73

RAW

-0.71

NetMessage

-0.71

rug

-0.71

udder

-0.67

hran

-0.66

olan

-0.65

si

-0.65

POSITIVE LOGITS

 exceed

0.91

0.87

ceed

0.84

 expectations

0.81

 exceeded

0.81

 exceeding

0.79

ingly

0.79

 saturation

0.76

atos

0.72

 amounts

0.71

Activations Density 0.008%

things that are more than a specified limit or standard

instances of the word "exceed" and its variations, indicating thresholds or limits being surpassed

No Comments

No Known Activations