INDEX

Explanations

superlatives, specifically words denoting the highest level of quality or achievement

oai_token-act-pair · gpt-3.5-turbo

superlative descriptors related to ranking or quality

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

 notice

-0.71

Nir

-0.65

 literal

-0.64

 territory

-0.64

 processing

-0.63

 reserved

-0.62

 manually

-0.61

 swoop

-0.61

 phased

-0.61

 congreg

-0.60

POSITIVE LOGITS

best

3.85

worst

2.34

Best

1.98

highest

1.75

better

1.64

 BEST

1.50

 best

1.49

most

1.47

 Best

1.44

perfect

1.35

Activations Density 0.016%

superlatives, specifically words denoting the highest level of quality or achievement

superlative descriptors related to ranking or quality

No Comments

No Known Activations