INDEX

Explanations

words indicating emphasis or importance like "worth" and "should"

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate the importance of noting or considering information

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ulous

-0.65

VID

-0.63

CoC

-0.60

ãĥı

-0.60

lite

-0.58

 Celt

-0.57

 Created

-0.56

ula

-0.56

 combinations

-0.56

POSITIVE LOGITS

 note

1.53

 noting

1.13

 mention

1.12

 emphas

1.09

 caveat

1.08

 underscore

1.05

note

1.05

 recall

1.05

 remember

1.04

 notice

1.04

Activations Density 0.135%

words indicating emphasis or importance like "worth" and "should"

phrases that indicate the importance of noting or considering information

No Comments

No Known Activations