INDEX

Explanations

names of individuals or officials in communication settings

oai_token-act-pair · gpt-3.5-turbo

names of individuals and entities involved in conversations or meetings

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

Pg

-0.69

 Dangerous

-0.65

 bets

-0.63

 profits

-0.62

 proportion

-0.62

 ILCS

-0.61

 dominates

-0.59

Quantity

-0.58

 compens

-0.58

 prev

-0.58

POSITIVE LOGITS

 representatives

0.92

 regarding

0.89

 Regarding

0.86

 thanking

0.81

 discussing

0.80

cerning

0.80

ilaterally

0.78

 representative

0.75

 asking

0.74

via

0.74

Activations Density 0.407%

names of individuals or officials in communication settings

names of individuals and entities involved in conversations or meetings

No Comments

No Known Activations