INDEX

Explanations

words related to the completion or finalization of tasks or events

oai_token-act-pair · gpt-3.5-turbo

instances of the word "end" or its variants in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

ichick

-0.68

ppo

-0.68

 underest

-0.68

IGHTS

-0.68

 Aval

-0.67

kaya

-0.67

ategory

-0.66

elsius

-0.63

edom

-0.62

 earthqu

-0.61

POSITIVE LOGITS

angered

1.00

angering

1.00

orph

0.99

urance

0.96

erer

0.89

lich

0.89

ering

0.86

ocrin

0.84

ulum

0.84

ancing

0.83

Activations Density 0.020%

words related to the completion or finalization of tasks or events

instances of the word "end" or its variants in various contexts

No Comments

No Known Activations