INDEX

Explanations

specific patterns followed by 'V'

oai_token-act-pair · gpt-3.5-turbo

occurrences of the letter "V"

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 understatement

-0.70

 scient

-0.63

 barley

-0.62

 bere

-0.61

istani

-0.60

 summons

-0.59

 upfront

-0.59

 doomed

-0.58

 dispers

-0.58

 pumpkin

-0.57

POSITIVE LOGITS

3.52

Vs

2.16

2.05

VT

1.98

VA

1.97

VI

1.93

VL

1.89

VB

1.87

1.86

VD

1.85

Activations Density 0.019%

specific patterns followed by 'V'

occurrences of the letter "V"

No Comments

No Known Activations