INDEX

Explanations

phrases related to forceful impact or criticism

oai_token-act-pair · gpt-3.5-turbo

instances of the word "slam" and its variations, indicating a focus on confrontational or critical actions

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

otype

-0.68

 Aware

-0.68

vernment

-0.67

ramer

-0.66

thus

-0.65

ensical

-0.64

earances

-0.64

lements

-0.64

 Lauder

-0.64

abet

-0.64

POSITIVE LOGITS

 dunk

1.01

 slammed

0.79

 hitters

0.78

 shut

0.78

 smack

0.77

 slams

0.77

onite

0.77

 down

0.76

 slamming

0.74

 goodbye

0.73

Activations Density 0.027%

phrases related to forceful impact or criticism

instances of the word "slam" and its variations, indicating a focus on confrontational or critical actions

No Comments

No Known Activations