INDEX

Explanations

questions starting with "Who" asking about various topics or actions

oai_token-act-pair · gpt-3.5-turbo

references to the word "who" in questions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

PORT

-0.70

strip

-0.64

 Pillar

-0.62

 saturation

-0.61

GV

-0.60

 spiral

-0.60

 Viking

-0.58

 Ashton

-0.58

interstitial

-0.57

BOX

-0.56

POSITIVE LOGITS

 cares

1.28

 else

1.16

 knows

1.16

soever

1.11

oping

1.09

ever

1.05

ops

1.04

oped

0.88

 knew

0.87

osh

0.86

Activations Density 0.027%

questions starting with "Who" asking about various topics or actions

references to the word "who" in questions

No Comments

No Known Activations