INDEX

Explanations

phrases related to instructional content

oai_token-act-pair · gpt-3.5-turbo

the phrase "how to" in various contexts

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

prem

-0.74

pict

-0.67

court

-0.66

 Coff

-0.65

cart

-0.64

 Porter

-0.64

personal

-0.64

CLUD

-0.64

enance

-0.64

 Appears

-0.62

POSITIVE LOGITS

 behave

1.14

 interpret

1.10

 treat

1.07

 proceed

1.05

 incorporate

1.05

 handle

1.04

 identify

1.03

 achieve

1.03

 navigate

1.02

 maximize

1.01

Activations Density 0.040%

phrases related to instructional content

the phrase "how to" in various contexts

No Comments

No Known Activations