INDEX

Explanations

references to physical rooms and locations

oai_token-act-pair · gpt-3.5-turbo

references to specific room numbers and their context

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

HER

-0.74

Uz

-0.69

Iss

-0.67

merce

-0.64

 Bonds

-0.63

pend

-0.63

elson

-0.62

usterity

-0.60

 retribution

-0.55

pleting

-0.55

POSITIVE LOGITS

room

1.08

rooms

0.94

 doors

0.93

 rooms

0.91

 upstairs

0.87

 room

0.86

lain

0.86

 divid

0.85

 occupancy

0.84

itory

0.84

Activations Density 0.060%

references to physical rooms and locations

references to specific room numbers and their context

No Comments

No Known Activations