INDEX

Explanations

phrases indicating communication of instructions, opinions, or information

oai_token-act-pair · gpt-3.5-turbo

expressions that involve the act of saying or communication

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

allery

-0.75

swick

-0.65

aylor

-0.63

 Areas

-0.61

ittens

-0.59

artments

-0.59

 Globe

-0.57

cdn

-0.57

edia

-0.57

oston

-0.56

POSITIVE LOGITS

 goodbye

0.98

 louder

0.81

 bluff

0.79

 loudly

0.76

 aloud

0.74

 Goodbye

0.69

ua

0.68

->

0.65

 politely

0.63

"#

0.62

Activations Density 0.328%

phrases indicating communication of instructions, opinions, or information

expressions that involve the act of saying or communication

No Comments

No Known Activations