INDEX

Explanations

names or titles at the start of passages

oai_token-act-pair · gpt-3.5-turbo

occurrences of the term "Per" and its variants

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

alogy

-0.70

 ladder

-0.65

 bets

-0.64

 fighters

-0.64

CBI

-0.64

 winding

-0.63

 headquarters

-0.63

 battling

-0.62

 elimination

-0.61

 diagonal

-0.60

POSITIVE LOGITS

Per

3.75

Per

2.25

per

1.90

PER

1.80

Pers

1.72

PER

1.54

Occ

1.18

Pen

1.17

per

1.16

perm

1.15

Activations Density 0.011%

names or titles at the start of passages

occurrences of the term "Per" and its variants

No Comments

No Known Activations