List your company

Question Answering on MedMCQA (test)

0.163Test Error RateGreedy MI
0.1620.168750.17550.18225Feb 8, 2026

Evaluation results

MethodLinks
Greedy MIk=42026.020.163————
Greedy MIk=52026.020.163————
Greedy MIk=32026.020.164————
Greedy MIk=62026.020.164————
Greedy MIk=72026.020.167————
Top-k (Acc)k=52026.020.17————
Top-k (Acc)k=72026.020.17————
Term 1 (Rel)k=72026.020.17————
Greedy MIk=82026.020.17————
Top-k (Acc)k=12026.020.171————
Term 1 (Rel)k=12026.020.171————
Terms 1+2 (mRMR)k=12026.020.171————
Greedy MIk=12026.020.171————
Terms 1+2 (mRMR)k=22026.020.171————
Greedy MIk=22026.020.171————
Terms 1+2 (mRMR)k=32026.020.171————
Terms 1+2 (mRMR)k=42026.020.171————
Top-k (Acc)k=62026.020.171————
Term 1 (Rel)k=62026.020.171————
Top-k (Acc)k=22026.020.172————
Term 1 (Rel)k=22026.020.172————
Term 1 (Rel)k=52026.020.172————
Top-k (Acc)k=82026.020.172————
Term 1 (Rel)k=82026.020.172————
Top-k (Acc)k=42026.020.173————
Term 1 (Rel)k=42026.020.173————
Terms 1+2 (mRMR)k=52026.020.173————
Term 1 (Rel)k=92026.020.173————
Top-k (Acc)k=32026.020.174————
Term 1 (Rel)k=32026.020.174————
Terms 1+2 (mRMR)k=62026.020.174————
Top-k (Acc)k=92026.020.174————
Greedy MIk=92026.020.174————
Terms 1+2 (mRMR)k=72026.020.175————
Terms 1+2 (mRMR)k=82026.020.176————
Terms 1+2 (mRMR)k=92026.020.177————
Top-k (Acc)k=102026.020.177————
Term 1 (Rel)k=102026.020.177————
Terms 1+2 (mRMR)k=102026.020.177————
Greedy MIk=102026.020.181————
Top-k (Acc)k=112026.020.182————
Term 1 (Rel)k=112026.020.182————
Greedy MIk=112026.020.183————
Terms 1+2 (mRMR)k=112026.020.184————
Top-k (Acc)k=122026.020.188————
Term 1 (Rel)k=122026.020.188————
Terms 1+2 (mRMR)k=122026.020.188————
Greedy MIk=122026.020.188————
CoG2026.02—27.55———
CriSPOLLM=Claude Instant, shot_count=0-shot2024.10—52.3———
CriSPOLLM=Claude Instant, shot_count=5-shot2024.10—54.4———
CriSPOLLM=Claude Sonnet, shot_count=0-shot2024.10—57.9———
CriSPOLLM=Claude Sonnet, shot_count=5-shot2024.10—57.4———
FoSS2026.02—27.44———
FullBackbone=Llama-3.1-8B, Training Data Percentage=100%, Data Selection Strategy=All data2026.04—40.53———
FullBackbone=Mistral-7B-v0.3, Training Data Percentage=100%, Data Selection Strategy=All data2026.04—36.55———
Kl-ModernBERT2026.06——36.7——
Kl-Nb-BERT2026.06——39.17——
Kl-NorBERT32026.06——34.15——
Llama-3-Med42-8BData Format=Plaintext2026.06—76.5176.5276.5776.51
Llama-3.1-8BData Format=Plaintext2026.06—70.2270.270.2670.22
Llama-3.1-Aloe-Beta-8BData Format=Plaintext2026.06—75.8775.8975.9475.87
ManualLLM=Claude Instant, shot_count=0-shot2024.10—49.2———
ManualLLM=Claude Instant, shot_count=5-shot2024.10—53.8———
ManualLLM=Claude Sonnet, shot_count=0-shot2024.10—49.8———
ManualLLM=Claude Sonnet, shot_count=5-shot2024.10—54.4———
Mistral-7B-Instruct-v0.3Data Format=Plaintext2026.06—63.5963.563.5663.59
ModernBERT2026.06——31.55——
Nb-BERT-large2026.06——28.44——
NorBERT3-large2026.06——32.51——
OPROLLM=Claude Instant2024.10—50.5———
OPROLLM=Claude Sonnet2024.10—57.7———
Qwen-2.5-7BData Format=Plaintext2026.06—71.9371.9472.0871.93
RandomBackbone=Llama-3.1-8B, Training Data Percentage=30%, Data Selection Strategy=Random sampling2026.04—32.32———
RandomBackbone=Mistral-7B-v0.3, Training Data Percentage=30%, Data Selection Strategy=Random sampling2026.04—33.11———
Transformer (w FT)fine-tuned=true, fine-tuning dataset=WikiText-1032026.02—26.95———
Transformer (w/o FT)fine-tuned=false2026.02—26.7———
wICIBackbone=Llama-3.1-8B, Training Data Percentage=30%, Data Selection Strategy=In-Context Influence2026.04—39.63———
wICIBackbone=Mistral-7B-v0.3, Training Data Percentage=30%, Data Selection Strategy=In-Context Influence2026.04—37.05———