INDEX

Explanations

attends to the first token in a sequence marked with "star" and the second token in a sequence marked with "bracket"

oai_attention-head · gpt-4o-mini Triggered by @bot

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-att/layer_0/width_16k/average_l0_61

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.0.attn.hook_z

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.02

1:0.01

2:0.01

3:0.01

4:0.01

5:0.01

6:0.07

7:0.06

8:0.03

9:0.02

10:0.01

11:0.01

12:0.61

13:0.02

14:0.01

15:0.02

Negative Logits

.)}

-0.82

")));

-0.80

}))

-0.79

 }}$}

-0.78

]")]

-0.78

})*/

-0.78

"]);

-0.78

]));

-0.77

}));

-0.77

}}}

-0.75

POSITIVE LOGITS

0.97

0.77

0.74

0.73

0.71

and

0.70

0.69

0.68

Activations Density 0.046%

attends to the first token in a sequence marked with "star" and the second token in a sequence marked with "bracket"

No Comments

No Known Activations