INDEX

Explanations

adjectives and nouns that are contextually related to a negative or critical sentiment

oai_token-act-pair · gpt-3.5-turbo

conjunctions and numbers that indicate a connection or comparison

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

":"/

-0.80

å§«

-0.78

ãĤ¼ãĤ¦ãĤ¹

-0.77

Ħ¢

-0.76

ãĥ¤

-0.75

NRS

-0.70

ãĥīãĥ©

-0.69

Enlarge

-0.69

pees

-0.66

acas

-0.65

POSITIVE LOGITS

 while

1.99

 whereas

1.96

 whilst

1.87

while

1.71

 followed

1.35

 respectively

1.27

 although

1.03

 Meanwhile

0.99

 versus

0.96

but

0.95

Activations Density 0.732%

adjectives and nouns that are contextually related to a negative or critical sentiment

conjunctions and numbers that indicate a connection or comparison

No Comments

No Known Activations