Canonik/lat-defuse-rag-repair

Pin Hugging Face revisions and record what was actually loaded 1 day ago

enhancement help wanted infra

1

Mechanistic evaluation of SAE-based activation interventions for correcting distractor-induced errors in RAG.

Python

Optimization forward: drop the KV cache, the full logits matrix, and the CPU hop 1 day ago

enhancement help wanted sae-gemma lat-defuse

Canonik/lat-defuse-rag-repair

1

Mechanistic evaluation of SAE-based activation interventions for correcting distractor-induced errors in RAG.

Python

Record where the pinned upstream diverges from the paper 1 day ago

documentation enhancement help wanted science

Canonik/lat-defuse-rag-repair

1

Mechanistic evaluation of SAE-based activation interventions for correcting distractor-induced errors in RAG.

Python

Optimization input: upstream never appends the shared answer prefix 1 day ago

bug help wanted lat-defuse science

Canonik/lat-defuse-rag-repair

1

Mechanistic evaluation of SAE-based activation interventions for correcting distractor-induced errors in RAG.

Python

Implement the stopping-rule tracker 1 day ago

good first issue lat-defuse

Canonik/lat-defuse-rag-repair

1

Mechanistic evaluation of SAE-based activation interventions for correcting distractor-induced errors in RAG.

Python

Implement the run-report writer 2 days ago

good first issue infra

Canonik/lat-defuse-rag-repair

1

Mechanistic evaluation of SAE-based activation interventions for correcting distractor-induced errors in RAG.

Python

Implement target selection and the optimization-input builder 3 days ago

good first issue science

Canonik/lat-defuse-rag-repair

1

Mechanistic evaluation of SAE-based activation interventions for correcting distractor-induced errors in RAG.

Python

Spike: survey datasets for the scaled-up study 3 days ago

good first issue dataset spike

Canonik/lat-defuse-rag-repair

1

Mechanistic evaluation of SAE-based activation interventions for correcting distractor-induced errors in RAG.

Python

Implement float32 logit metrics 3 days ago

good first issue science

Canonik/lat-defuse-rag-repair

1

Mechanistic evaluation of SAE-based activation interventions for correcting distractor-induced errors in RAG.

Python

Add the SAE compatibility check 3 days ago

good first issue sae-gemma

Canonik/lat-defuse-rag-repair

1

Mechanistic evaluation of SAE-based activation interventions for correcting distractor-induced errors in RAG.

Python

Fill in the experiment config object and shared error type 5 days ago

good first issue infra

Canonik/lat-defuse-rag-repair

1

Mechanistic evaluation of SAE-based activation interventions for correcting distractor-induced errors in RAG.

Python

Open Issues Need Help