INDEX

Explanations

names and descriptions of individuals involved in various actions or situations

oai_token-act-pair · gpt-3.5-turbo

instances of the word "who," which likely indicates a focus on discussing individuals or characters

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

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

ulp

-0.70

CONCLUS

-0.69

Something

-0.63

Conclusion

-0.61

âĵĺ

-0.60

Untitled

-0.59

tnc

-0.59

 causation

-0.59

agine

-0.59

 Directorate

-0.57

POSITIVE LOGITS

 owns

1.17

 suffers

1.09

 grew

1.08

 earns

1.07

 underwent

1.06

 resides

1.04

 lives

1.04

 wears

1.03

 died

1.01

 prefers

1.01

Activations Density 0.085%

names and descriptions of individuals involved in various actions or situations

instances of the word "who," which likely indicates a focus on discussing individuals or characters

No Comments

No Known Activations