List your company

Faithfulness Evaluation on Halogen

73CODEDistributional Semantics Tracing
22.0435.2748.561.73Oct 7, 2025

Evaluation results

MethodLinks
Distributional Semantics TracingModel=Gemma2-9B2025.10737783847879
Sparse AutoencodersModel=Gemma2-9B2025.10645361526359
Distributional Semantics TracingModel=Gemma2-2B2025.10635967776566
PatchscopesModel=Gemma2-9B2025.10626845644958
Subsequence Analysis TracingModel=Gemma2-2B2025.10574556555954
Causal Path TracingModel=Gemma2-2B2025.10565465544956
Causal Path TracingModel=Gemma2-9B2025.10565460655658
Sparse AutoencodersModel=Gemma2-2B2025.10544951445551
Token Evolution (Logit Lens)Model=Gemma2-9B2025.10544345455448
Subsequence Analysis TracingModel=Gemma2-9B2025.10545956575957
PatchscopesModel=Gemma2-2B2025.10495653455451
Token Evolution (Logit Lens)Model=Gemma2-2B2025.10455244543947
attentionModel=Gemma2-9B2025.10452931212230
ReagentModel=Gemma2-9B2025.10424443453341
ReagentModel=Gemma2-2B2025.10354519241929
limeModel=Gemma2-9B2025.10343345292934
gradient-shapModel=Gemma2-9B2025.10324035342032
gradient-shapModel=Gemma2-2B2025.10313513101521
limeModel=Gemma2-2B2025.10293215131421
attentionModel=Gemma2-2B2025.10242819121219