INDEX

Explanations

phrases that contrast or add additional information to the preceding text

oai_token-act-pair · gpt-3.5-turbo

references to conditional statements or clauses

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

Rac

-0.79

 PLUS

-0.76

chwitz

-0.67

ãĤ´

-0.66

lette

-0.65

ãĢĤ

-0.61

Yok

-0.61

onds

-0.61

 Indra

-0.61

 Chevron

-0.60

POSITIVE LOGITS

 technically

1.13

 admittedly

0.84

 differ

0.84

 differed

0.83

 initially

0.82

 somewhat

0.79

 theoretically

0.79

 undeniably

0.78

 disagree

0.77

 differs

0.76

Activations Density 0.399%

phrases that contrast or add additional information to the preceding text

references to conditional statements or clauses

No Comments

No Known Activations