INDEX

Explanations

phrases indicating incomplete actions or tasks

oai_token-act-pair · gpt-3.5-turbo

phrases indicating actions that are yet to be completed or achieved

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

Natural

-0.69

 Rite

-0.67

Nap

-0.67

 Measures

-0.65

quickShipAvailable

-0.64

 Compass

-0.64

BAS

-0.63

Alternative

-0.63

Loop

-0.62

mens

-0.62

POSITIVE LOGITS

 officially

0.97

 confirm

0.93

 confir

0.90

 disappoint

0.89

 settle

0.88

 sustain

0.84

 disclose

0.84

 fulfill

0.84

 mention

0.83

 acknow

0.82

Activations Density 0.057%

phrases indicating incomplete actions or tasks

phrases indicating actions that are yet to be completed or achieved

No Comments

No Known Activations