List your company

Multihop Question Answering on MuSiQue (EM, ASR)

42EMPARADOX
0.411.22232.8Jul 1, 2026

Evaluation results

MethodLinks
PARADOXFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07423
Clean (No Attack)Framework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.0738—
PARADOXFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.07342
Clean (No Attack)Framework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.0732—
NaiveFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.073113
TopicFramework=WebThinker, LLM Backbone=QwQ-32B2026.07300
PoisonedRAGFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07281
PARADOXFramework=WebThinker, LLM Backbone=QwQ-32B2026.07272
Fake CompletionFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07261
NaiveFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07262
IgnoreFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.072517
PARADOXFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07240
Clean (No Attack)Framework=WebThinker, LLM Backbone=QwQ-32B2026.0723—
IgnoreFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07219
NaiveFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.072030
Fake CompletionFramework=WebThinker, LLM Backbone=QwQ-32B2026.072011
CombinedFramework=WebThinker, LLM Backbone=QwQ-32B2026.07195
IgnoreFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07192
NaiveFramework=WebThinker, LLM Backbone=QwQ-32B2026.071811
PoisonedRAGFramework=WebThinker, LLM Backbone=QwQ-32B2026.07180
Fake CompletionFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07182
Fake CompletionFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.071745
PoisonedRAGFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.07172
TopicFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07170
Clean (No Attack)Framework=WebThinker, LLM Backbone=DeepSeek-32B2026.0717—
CombinedFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07160
TopicFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07160
PoisonedRAGFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07160
IgnoreFramework=WebThinker, LLM Backbone=QwQ-32B2026.071510
CombinedFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.071444
CombinedFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07141
TopicFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.071318
KidnapRAGFramework=WebThinker, LLM Backbone=QwQ-32B2026.07930
KidnapRAGFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07930
KidnapRAGFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.07866
KidnapRAGFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07260