INDEX

Explanations

terms related to reviews or evaluations, especially when associated with specific names or titles

oai_token-act-pair · gpt-3.5-turbo

references to "Review"

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

 enslaved

-0.76

 fort

-0.72

 sunk

-0.65

veh

-0.65

 forced

-0.64

 marching

-0.64

 stab

-0.63

 ensl

-0.63

dom

-0.62

flu

-0.62

POSITIVE LOGITS

 Review

4.02

review

2.55

Review

2.47

 review

2.33

 Reviews

2.31

 reviews

1.96

 reviewer

1.87

Reviewed

1.83

 reviewers

1.68

 reviewed

1.64

Activations Density 0.018%

terms related to reviews or evaluations, especially when associated with specific names or titles

references to "Review"

No Comments

No Known Activations