INDEX

Explanations

references to e-readers and related concepts

oai_token-act-pair · gpt-3.5-turbo

terms related to reading and literature

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

MD

-0.76

 honor

-0.75

cho

-0.67

 forfe

-0.66

 carb

-0.65

JPM

-0.64

 salt

-0.63

 penal

-0.62

 Ranch

-0.62

 mega

-0.62

POSITIVE LOGITS

reading

3.86

reads

2.17

reader

1.99

telling

1.70

Reading

1.62

watching

1.38

writing

1.33

playing

1.28

finding

1.18

checking

1.18

Activations Density 0.025%

references to e-readers and related concepts

terms related to reading and literature

No Comments

No Known Activations