INDEX

Explanations

phrases relating to understanding or questioning how something works or is implemented

oai_token-act-pair · gpt-3.5-turbo

references to a subject or concept, specifically the word "it" and its variations in context

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

allery

-0.73

teness

-0.70

 Frazier

-0.69

ãĥĥ

-0.67

ãĥĩãĤ£

-0.66

 Bliss

-0.62

ANCE

-0.61

chens

-0.61

anova

-0.60

rano

-0.60

POSITIVE LOGITS

 interpreted

0.82

 unfolded

0.76

 perce

0.73

 intersect

0.73

 stacked

0.72

 behaved

0.71

 perspect

0.70

 hurd

0.70

 interpret

0.69

'd

0.69

Activations Density 0.195%

phrases relating to understanding or questioning how something works or is implemented

references to a subject or concept, specifically the word "it" and its variations in context

No Comments

No Known Activations