INDEX

Explanations

dates and numerical values

oai_token-act-pair · gpt-3.5-turbo

numeric values and special characters

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 conservatives

-0.62

 tweaked

-0.62

DJs

-0.60

 Robot

-0.60

 Dele

-0.60

 broadcast

-0.59

 strategically

-0.59

 disruptive

-0.59

 activists

-0.58

 targeted

-0.58

POSITIVE LOGITS

Â«

1.12

^^

1.03

 âĸł

0.95

0.94

Si

0.93

Â»

0.90

Â£

0.88

 tion

0.87

 Â»

0.86

sic

0.85

Activations Density 0.151%

dates and numerical values

numeric values and special characters

No Comments

No Known Activations