List your company

Question Answering on NQ (ACCR and ACCL)

87AccuracyHRAG
42.2853.8965.577.11May 15, 2025Jul 15, 2025Sep 15, 2025Nov 16, 2025Jan 17, 2026Mar 20, 2026May 21, 2026

Evaluation results

MethodLinks
HRAGModel=Llama3-70B, Defense Framework=AstuteRAG2025.0587——5—
CleanModel=Llama3-70B, Defense Framework=AstuteRAG2025.0586————
PRAGModel=Deepseek-V3-0324, Defense Framework=AstuteRAG2025.0586——0—
CleanModel=Deepseek-V3-0324, Defense Framework=AstuteRAG2025.0586————
CleanModel=GPT-4o, Defense Framework=AstuteRAG2025.0584————
HRAGModel=Deepseek-V3-0324, Defense Framework=AstuteRAG2025.0584——6—
PRAGModel=GPT-4o, Defense Framework=InstructRAG2025.0583——14—
HRAGModel=GPT-4, Defense Framework=AstuteRAG2025.0583——7—
PRAGModel=Llama3-70B, Defense Framework=AstuteRAG2025.0583——5—
CleanModel=Deepseek-V3-0324, Defense Framework=InstructRAG2025.0582————
HRAGModel=GPT-4o, Defense Framework=AstuteRAG2025.0582——9—
PRAGModel=GPT-4, Defense Framework=AstuteRAG2025.0581——5—
CleanModel=GPT-4, Defense Framework=AstuteRAG2025.0580————
PRAGModel=GPT-4o, Defense Framework=AstuteRAG2025.0580——8—
CleanModel=Llama3-70B, Defense Framework=InstructRAG2025.0579————
PRAGModel=Deepseek-V3-0324, Defense Framework=InstructRAG2025.0579——21—
CleanModel=GPT-4o, Defense Framework=InstructRAG2025.0578————
PRAGModel=Llama3-8B, Defense Framework=AstuteRAG2025.0578——9—
CleanModel=Llama3-8B, Defense Framework=AstuteRAG2025.0578————
CleanModel=GPT-4, Defense Framework=InstructRAG2025.0575————
HRAGModel=GPT-4o, Defense Framework=InstructRAG2025.0575——18—
HRAGModel=GPT-3.5, Defense Framework=AstuteRAG2025.0575——9—
AuthChainModel=Deepseek-V3-0324, Defense Framework=AstuteRAG2025.0575——11—
CleanModel=Llama3-8B, Defense Framework=InstructRAG2025.0574————
DCRDModel=Llama3-8B2026.0573.4————
HRAGModel=Llama3-70B, Defense Framework=InstructRAG2025.0572——16—
DCRDModel=Llama2-13B2026.0571.4————
CleanModel=GPT-3.5, Defense Framework=AstuteRAG2025.0571————
AuthChainModel=GPT-4o, Defense Framework=AstuteRAG2025.0571——23—
AuthChainModel=GPT-4o, Defense Framework=InstructRAG2025.0569——27—
HRAGModel=Deepseek-V3-0324, Defense Framework=InstructRAG2025.0569——28—
AuthChainModel=GPT-4, Defense Framework=AstuteRAG2025.0569——23—
HRAGModel=Llama3-8B, Defense Framework=AstuteRAG2025.0569——17—
COIECDModel=Llama2-13B2026.0568.6————
DCRDModel=Llama2-7B2026.0568.4————
COIECDModel=Llama3-8B2026.0568.4————
ADACADModel=Llama2-13B2026.0568.2————
AuthChainModel=Deepseek-V3-0324, Defense Framework=InstructRAG2025.0568——30—
GreedyModel=Llama3-8B2026.0567.4————
CleanModel=GPT-3.5, Defense Framework=InstructRAG2025.0566————
ADACADModel=Llama2-7B2026.0565.4————
ADACADModel=Llama3-8B2026.0565.2————
PRAGModel=Llama3-70B, Defense Framework=InstructRAG2025.0565——30—
GreedyModel=Llama2-13B2026.0564.3————
PRAGModel=GPT-4, Defense Framework=InstructRAG2025.0564——29—
RADARTop-k=10, Pos=Pos 10, Generator LLM=Grok-4-fast2026.0562.4——5—
PRAGModel=GPT-3.5, Defense Framework=AstuteRAG2025.0562——21—
ReliabilityRAGTop-k=50, Pos=Pos 25, Generator LLM=Grok-4-fast2026.0562——3.6—
RADARTop-k=50, Pos=Pos 25, Generator LLM=Grok-4-fast2026.0562——4—
RobustRAGTop-k=50, Pos=Pos 50, Generator LLM=Grok-4-fast2026.0562——3.2—
ReliabilityRAGTop-k=50, Pos=Pos 50, Generator LLM=Grok-4-fast2026.0562——4—
RADARTop-k=50, Pos=Pos 50, Generator LLM=Grok-4-fast2026.0562——4—
ReliabilityRAGTop-k=10, Pos=Pos 10, Generator LLM=Grok-4-fast2026.0561.4——9.8—
RobustRAGTop-k=50, Pos=Pos 25, Generator LLM=Grok-4-fast2026.0561.2——2.8—
RADARTop-k=50, Pos=Pos 1, Generator LLM=Grok-4-fast2026.0561——4.8—
DCRDModel=Mistral-7B2026.0560.9————
AuthChainModel=Llama3-8B, Defense Framework=AstuteRAG2025.0560——33—
COIECDModel=Llama2-7B2026.0559.9————
RADARTop-k=10, Pos=Pos 1, Generator LLM=Grok-4-fast2026.0559.4——8.2—
COIECDModel=Mistral-7B2026.0559.1————
GreedyModel=Mistral-7B2026.0558.8————
ADACADModel=Mistral-7B2026.0558.1————
AuthChainModel=Llama3-70B, Defense Framework=AstuteRAG2025.0558——38—
RobustRAGTop-k=50, Pos=Pos 1, Generator LLM=Grok-4-fast2026.0558——5.2—
ReliabilityRAGTop-k=10, Pos=Pos 1, Generator LLM=Grok-4-fast2026.0557——16.8—
RobustRAGTop-k=10, Pos=Pos 10, Generator LLM=Grok-4-fast2026.0557——2.2—
ILRBackbone=Llama-3.1-8B-Instruct, Training Group=Group 22025.0955.31————
ReliabilityRAGTop-k=50, Pos=Pos 1, Generator LLM=Grok-4-fast2026.0555——25.4—
ILRBackbone=Llama-3.1-8B-Instruct, Training Group=Group 12025.0954.77————
RobustRAGTop-k=10, Pos=Pos 1, Generator LLM=Grok-4-fast2026.0554.4——5.2—
HRAGModel=GPT-4, Defense Framework=InstructRAG2025.0554——40—
PRAGModel=Llama3-8B, Defense Framework=InstructRAG2025.0554——38—
ILRBackbone=Qwen2.5-14B-Instruct, Training Group=Group 32025.0953.87————
CADModel=Mistral-7B2026.0552.6————
ILRBackbone=Qwen2.5-14B-Instruct, Training Group=Group 22025.0952.45————
AstuteRAGTop-k=10, Pos=Pos 10, Generator LLM=Grok-4-fast2026.0552.4——2.6—
DPOBackbone=Llama-3.1-8B-Instruct2025.0952.32————
AstuteRAGTop-k=10, Pos=Pos 1, Generator LLM=Grok-4-fast2026.0552.2——2.6—
GRPOBackbone=Llama-3.1-8B-Instruct2025.0952.11————
GreedyModel=Llama2-7B2026.0551.9————
InstructRAGTop-k=50, Pos=Pos 25, Generator LLM=Grok-4-fast2026.0551.6——3.6—
CADModel=Llama2-13B2026.0551.3————
BaseBackbone=Llama-3.1-8B-Instruct2025.0951.12————
AuthChainModel=GPT-4, Defense Framework=InstructRAG2025.0551——44—
CADModel=Llama3-8B2026.0551————
InstructRAGTop-k=50, Pos=Pos 50, Generator LLM=Grok-4-fast2026.0550.4——4—
CADModel=Llama2-7B2026.0550.3————
HRAGModel=Llama3-8B, Defense Framework=InstructRAG2025.0550——40—
GRPOBackbone=Qwen2.5-14B-Instruct2025.0950————
AstuteRAGTop-k=50, Pos=Pos 25, Generator LLM=Grok-4-fast2026.0550——3.8—
AstuteRAGTop-k=50, Pos=Pos 50, Generator LLM=Grok-4-fast2026.0549.8——3—
DPOBackbone=Qwen2.5-14B-Instruct2025.0949.6————
AstuteRAGTop-k=50, Pos=Pos 1, Generator LLM=Grok-4-fast2026.0549.2——2.2—
BaseBackbone=Qwen2.5-14B-Instruct2025.0949.13————
PRAGModel=GPT-3.5, Defense Framework=InstructRAG2025.0549——42—
AuthChainModel=Llama3-8B, Defense Framework=InstructRAG2025.0549——44—
InstructRAGTop-k=10, Pos=Pos 10, Generator LLM=Grok-4-fast2026.0548.6——5.4—
AuthChainModel=Llama3-70B, Defense Framework=InstructRAG2025.0547——45—
HRAGModel=GPT-3.5, Defense Framework=InstructRAG2025.0545——50—
AuthChainModel=GPT-3.5, Defense Framework=AstuteRAG2025.0544——47—