INDEX

Explanations

information on news stories that readers might find important or relevant

oai_token-act-pair · gpt-3.5-turbo

references to news stories

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

EStream

-0.76

wik

-0.69

ucl

-0.68

=-=-=-=-

-0.66

oute

-0.65

 acquaintance

-0.64

ersen

-0.64

;;;;;;;;;;;;

-0.64

 relocation

-0.60

 separation

-0.58

POSITIVE LOGITS

mith

0.82

eming

0.70

linger

0.69

pace

0.68

 Stories

0.68

EAR

0.67

poon

0.65

hel

0.65

cale

0.65

IDER

0.65

Activations Density 0.050%

information on news stories that readers might find important or relevant

references to news stories

No Comments

No Known Activations