INDEX

Explanations

phrases indicating a call to action or a leadership role

oai_token-act-pair · gpt-3.5-turbo

phrases indicating uniqueness or being the "first" or "best" in a context

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

mares

-0.77

aired

-0.76

Ö¼

-0.71

 indefinitely

-0.66

 extensively

-0.64

 aimed

-0.63

 preceded

-0.63

 followed

-0.62

cies

-0.61

 cylinders

-0.61

POSITIVE LOGITS

 happiest

0.97

 bearer

0.96

 smartest

0.92

 Chosen

0.91

 beneficiary

0.90

guy

0.87

 fitt

0.87

 underdog

0.86

 slightest

0.83

 careg

0.82

Activations Density 0.151%

phrases indicating a call to action or a leadership role

phrases indicating uniqueness or being the "first" or "best" in a context

No Comments

No Known Activations