INDEX

Explanations

specific consecutive characters or character sequences

oai_token-act-pair · gpt-3.5-turbo

punctuation marks signaling significant statements or questions

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

æĪ¦

-0.65

itton

-0.64

 earthqu

-0.63

 pione

-0.60

 teasp

-0.59

apsed

-0.59

éĹĺ

-0.58

 unintention

-0.57

 unnecess

-0.57

cloneembedreportprint

-0.57

POSITIVE LOGITS

 Does

1.34

Are

1.25

How

1.24

 does

1.24

does

1.23

 Would

1.23

Is

1.20

Does

1.19

 Could

1.17

Did

1.17

Activations Density 0.545%

specific consecutive characters or character sequences

punctuation marks signaling significant statements or questions

No Comments

No Known Activations