INDEX

Explanations

mathematical symbols and expressions

oai_token-act-pair · gpt-3.5-turbo

special characters or punctuation marks frequently used in code contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

gag

-0.71

 bast

-0.66

 Franch

-0.61

 date

-0.61

 belongs

-0.60

 experiment

-0.60

 belonging

-0.60

 belong

-0.60

 fuse

-0.59

 thunder

-0.59

POSITIVE LOGITS

¸

4.40

´

2.32

·

2.12

¹

2.08

¼

2.07

°

2.02

¶

1.97

¾

1.93

º

1.92

µ

1.90

Activations Density 0.009%

mathematical symbols and expressions

special characters or punctuation marks frequently used in code contexts

No Comments

No Known Activations