INDEX

Explanations

strongly negative or critical language

oai_token-act-pair · gpt-3.5-turbo

instances of the word "blasted" and its variations used in a critical context

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

slave

-0.95

abet

-0.84

adra

-0.78

thus

-0.76

Georg

-0.74

estate

-0.74

ffic

-0.72

Interstitial

-0.71

Solution

-0.71

vent

-0.70

POSITIVE LOGITS

 blasted

1.13

 blasting

1.13

 blasts

0.96

umenthal

0.91

abwe

0.83

 barr

0.81

nesday

0.78

onite

0.77

 stares

0.71

anked

0.70

Activations Density 0.012%

strongly negative or critical language

instances of the word "blasted" and its variations used in a critical context

No Comments

No Known Activations