INDEX

Explanations

words related to contrasting or negating statements

oai_token-act-pair · gpt-3.5-turbo

conjunctions and transition phrases in discourse

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

 Objects

-0.65

uggest

-0.64

 chips

-0.60

 lump

-0.59

 grains

-0.59

 rods

-0.58

 overboard

-0.58

terness

-0.57

 Heads

-0.57

 slate

-0.57

POSITIVE LOGITS

unia

0.80

 albeit

0.79

um

0.79

uh

0.75

but

0.75

andom

0.73

 except

0.68

necess

0.68

yet

0.68

except

0.67

Activations Density 0.230%

words related to contrasting or negating statements

conjunctions and transition phrases in discourse

No Comments

No Known Activations