INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L5_Hcat_z_lr1.20e-03_l11.00e+00_ds49152_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

float32

Hook Name

blocks.5.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.08

1:0.06

2:0.08

3:0.08

4:0.09

5:0.10

6:0.08

7:0.06

8:0.08

9:0.09

10:0.08

11:0.07

Negative Logits

 resil

-1.76

.","

-1.69

 Imperium

-1.61

Merit

-1.55

 stewards

-1.53

�

-1.51

atform

-1.49

 keynote

-1.48

Physical

-1.48

 initi

-1.48

POSITIVE LOGITS

�

1.70

cv

1.60

luaj

1.60

 botched

1.51

 Duffy

1.50

cia

1.50

�

1.47

 tampering

1.46

named

1.46

 Clip

1.44

Activations Density 0.000%

No Known Activations

This feature has no known activations.