INDEX

Explanations

names mentioned in a news context

oai_token-act-pair · gpt-3.5-turbo

mentions of specific individuals or characters in a news context

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Compatibility

-0.64

ishing

-0.63

ish

-0.63

Different

-0.61

oline

-0.58

ãĥ¥

-0.57

":"/

-0.56

zing

-0.56

=#

-0.55

irie

-0.55

POSITIVE LOGITS

 whom

1.15

who

1.15

Jr

1.08

who

1.06

 whose

1.04

whose

0.99

aka

0.94

 resigned

0.88

 withdrew

0.87

 died

0.85

Activations Density 0.158%

names mentioned in a news context

mentions of specific individuals or characters in a news context

No Comments

No Known Activations