INDEX

Explanations

the word "that" coupled with a strong emotional connotation

oai_token-act-pair · gpt-3.5-turbo

negations, particularly the contraction "doesn't" and its variations

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 spirited

-0.74

 Returning

-0.72

eous

-0.69

 Maiden

-0.67

 Laun

-0.64

 Galile

-0.63

 Empires

-0.62

 discovering

-0.61

iscovered

-0.60

 Presence

-0.60

POSITIVE LOGITS

 matter

1.30

 mean

1.16

matter

1.16

 preclude

1.02

 mattered

1.01

 negate

0.99

 necessarily

0.99

 constitute

0.96

mean

0.96

necess

0.96

Activations Density 0.074%

the word "that" coupled with a strong emotional connotation

negations, particularly the contraction "doesn't" and its variations

No Comments

No Known Activations