INDEX

Explanations

words associated with fraud or deceit

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

New Auto-Interp

Configuration

jbloom/GPT2-Small-OAI-v5-128k-attn-out-SAEs/v5_128k_layer_0

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

131,072

Data Type

torch.float32

Hook Name

blocks.0.hook_attn_out

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ãĤ¼ãĤ¦ãĤ¹

-0.92

å¥³

-0.79

 Leilan

-0.79

ipher

-0.78

\\\\\\\\

-0.78

çīĪ

-0.77

ANC

-0.76

illin

-0.73

ÃŃn

-0.73

ansk

-0.72

POSITIVE LOGITS

posts

0.77

bered

0.77

VERT

0.73

bers

0.70

ammed

0.70

pees

0.68

bag

0.67

vert

0.67

bags

0.66

cam

0.65

Activations Density 0.002%

words associated with fraud or deceit

No Comments

No Known Activations