EAP / EAP-IG circuit discovery

Edge Attribution Patching (EAP) or EAP with Integrated Gradients (EAP-IG) discovers safety-relevant edges in the model's computational graph.

from safetune.interpret import eap_safety_circuit, EAPSafetyCircuitConfig

# NOTE: eap_safety_circuit takes a HuggingFace model ID string, not a model object.
# EAP requires repeated clean/corrupted forward passes with precise activation caching;
# it loads and manages its own model copy internally to avoid interference with hooks
# on an already-loaded model.
circuit = eap_safety_circuit(
    "meta-llama/Llama-3.2-3B-Instruct",  # HF model ID (string)
    harmful_prompts=harmful,
    harmless_prompts=harmless,
    config=EAPSafetyCircuitConfig(
        method="eap-ig",           # "eap" or "eap-ig"
        granularity="head",        # "head" or "block"
        top_k_edges=100,
    ),
)

# eap_safety_circuit returns the same CircuitInfo shape as safety_circuit_info():
print(len(circuit.safety_units.unit_ids), "safety-relevant edges/heads found")
print(circuit.safety_units.unit_ids[:5])

EAPSafetyCircuitConfig

Field	Type	Default	Description
method	str	"eap-ig"	"eap" or "eap-ig"
granularity	str	"head"	"head" or "block"
intervention	str	"patching"	"patching", "zero", or "mean"
top_k_edges	int	100	Number of edges to keep
ig_steps	int	5	Integration steps for EAP-IG
batch_size	int	8	Batch size
max_seq_len	int	64	Max sequence length

When to use

EAP discovers the circuit (set of edges) responsible for refusal, not just individual neurons. Use when you need to understand the full computation path that drives the model's safety behaviour.

Citations

@article{eap2023,
  title  = {Attribution Patching Outperforms Automated Circuit Discovery},
  author = {Syed, Aaquib and Rager, Can and Conmy, Arthur},
  year   = {2023},
  note   = {NeurIPS 2023 ATTRIB Workshop, arXiv:2310.10348},
}

@article{eapig2024,
  title  = {Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanisms},
  author = {Hanna, Michael and Pezzelle, Sandro and Belinkov, Yonatan},
  year   = {2024},
  note   = {COLM 2024, arXiv:2403.17806},
}