INDEX

Explanations

references to deception or manipulation in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

torch.float32

Hook Point

blocks.8.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

detail

-0.81

hani

-0.71

ailability

-0.71

ĪĴ

-0.70

ģ«

-0.67

namese

-0.64

regon

-0.63

idav

-0.61

icient

-0.61

enable

-0.60

POSITIVE LOGITS

 himself

0.81

liest

0.75

 feared

0.73

 Himself

0.72

,—

0.67

 yourselves

0.66

ghazi

0.66

 fame

0.66

 itself

0.65

 originally

0.65

Activations Density 0.230%

references to deception or manipulation in various contexts

No Comments

No Known Activations