INDEX

Explanations

timestamps in a format "+X" where X is a number, possibly related to progress or enhancements

oai_token-act-pair · gpt-3.5-turbo

occurrences of the character '+'

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

opol

-0.68

 quotation

-0.64

Mr

-0.64

odic

-0.63

 repr

-0.62

ouri

-0.61

 ï¿½

-0.60

 conspiracy

-0.60

 caric

-0.58

''

-0.57

POSITIVE LOGITS

3.75

+.

2.97

+,

2.95

+)

2.89

+(

2.04

++

2.01

+++

1.94

1.89

-+

1.83

)+

1.67

Activations Density 0.006%

timestamps in a format "+X" where X is a number, possibly related to progress or enhancements

occurrences of the character '+'

No Comments

No Known Activations