INDEX

Explanations

links between actions and punishments, particularly involving suspensions

oai_token-act-pair · gpt-3.5-turbo

mentions of suspension in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

zig

-0.79

ebus

-0.78

ipped

-0.76

rics

-0.75

illard

-0.73

åĮ

-0.69

opher

-0.68

swick

-0.67

 originals

-0.66

phies

-0.66

POSITIVE LOGITS

 indefinitely

1.17

 disbelief

1.06

 suspensions

0.95

 suspending

0.95

 suspended

0.92

 suspension

0.90

ishment

0.90

 suspend

0.81

 probation

0.75

ension

0.74

Activations Density 0.030%

links between actions and punishments, particularly involving suspensions

mentions of suspension in various contexts

No Comments

No Known Activations