INDEX

Explanations

names of individuals, potentially focusing on people making specific announcements or engaging in activities

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

ĺħ

-0.67

Translation

-0.65

Conclusion

-0.59

])

-0.59

 anymore

-0.58

urg

-0.57

endif

-0.57

ĪĴ

-0.57

Fuck

-0.56

¼

-0.56

POSITIVE LOGITS

 joins

1.19

 joined

1.12

 arrived

1.10

 graduated

1.08

 tweeted

1.06

 underwent

1.05

 debuted

1.05

 told

1.01

 reportedly

1.01

 appeared

0.98

Activations Density 0.222%

names of individuals, potentially focusing on people making specific announcements or engaging in activities

No Comments

No Known Activations