INDEX

Explanations

expressions or mentions of intentions or purposes

oai_token-act-pair · gpt-3.5-turbo

the word "intended" 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 @ 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

poll

-0.65

ucket

-0.60

oner

-0.59

phi

-0.58

asia

-0.58

Trend

-0.57

pathy

-0.57

rank

-0.57

 Anarchy

-0.56

$$

-0.56

POSITIVE LOGITS

 intended

3.74

 meant

1.86

 intending

1.77

 designed

1.76

 planned

1.69

 aimed

1.69

 intend

1.62

 envisioned

1.55

 hoped

1.50

 intention

1.44

Activations Density 0.010%

expressions or mentions of intentions or purposes

the word "intended" in various contexts

No Comments

No Known Activations