INDEX

Explanations

keywords related to drawing attention to specific information

oai_token-act-pair · gpt-3.5-turbo

instances of the word "pointed" or its variations, indicating references to observations or assertions made by individuals

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

DAQ

-0.88

uthor

-0.79

namese

-0.72

ccording

-0.70

natureconservancy

-0.69

eco

-0.69

ingham

-0.68

ibaba

-0.68

iverpool

-0.67

estern

-0.66

POSITIVE LOGITS

 pointing

0.98

 pointed

0.93

lessly

0.85

 blank

0.82

pointers

0.73

 finger

0.72

 toward

0.72

points

0.72

 deductions

0.71

 fingers

0.70

Activations Density 0.029%

keywords related to drawing attention to specific information

instances of the word "pointed" or its variations, indicating references to observations or assertions made by individuals

No Comments

No Known Activations