INDEX

Explanations

expressions of empathy or sympathy

oai_token-act-pair · gpt-3.5-turbo

expressions of apology or regret

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

minecraft

-0.79

sports

-0.77

ouver

-0.73

tein

-0.71

rouse

-0.70

hack

-0.68

tnc

-0.68

authorized

-0.67

craft

-0.66

impro

-0.64

POSITIVE LOGITS

 sorry

1.37

 Sorry

1.00

Sorry

0.94

sorry

0.94

 excuse

0.88

 pardon

0.84

 apologies

0.78

soever

0.77

 THANK

0.76

à¤¾

0.76

Activations Density 0.007%

expressions of empathy or sympathy

expressions of apology or regret

No Comments

No Known Activations