INDEX

Explanations

phrases containing instructions or warnings

oai_token-act-pair · gpt-3.5-turbo

phrases that convey warnings or advice against certain actions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

ãĤ¦ãĤ¹

-0.81

quickShipAvailable

-0.80

 proving

-0.68

 Reborn

-0.68

albeit

-0.68

 assurances

-0.66

 Crusade

-0.65

 Preservation

-0.63

 assures

-0.63

omen

-0.62

POSITIVE LOGITS

 bother

1.23

 underestimate

1.15

 interfere

1.13

 hesitate

1.08

 worry

1.07

 confuse

1.05

 medd

1.03

 disturb

1.00

 discriminate

0.99

 stray

0.99

Activations Density 0.174%

phrases containing instructions or warnings

phrases that convey warnings or advice against certain actions

No Comments

No Known Activations