INDEX

Explanations

instances where an action or outcome is typically expected or occurs most often

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate frequency or common occurrences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 Orchestra

-0.71

 Agenda

-0.70

IDA

-0.66

eem

-0.66

 Abdullah

-0.66

Kob

-0.65

 Feld

-0.64

 Shard

-0.64

ultan

-0.62

 Integrity

-0.62

POSITIVE LOGITS

entimes

1.03

 overlooked

0.87

 regarded

0.82

 frowned

0.82

 referred

0.81

 relegated

0.81

 reserved

0.78

 accomp

0.78

 consist

0.78

 suspects

0.76

Activations Density 0.036%

instances where an action or outcome is typically expected or occurs most often

phrases that indicate frequency or common occurrences

No Comments

No Known Activations