List your company

Multihop Question Answering on 2WikiMultihopQA (EM, ASR)

60EMClean (No Attack)
6.9620.7334.548.27Jul 1, 2026

Evaluation results

MethodLinks
Clean (No Attack)Framework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.0760—
Clean (No Attack)Framework=WebThinker, LLM Backbone=DeepSeek-32B2026.0758—
PARADOXFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07570
Clean (No Attack)Framework=WebThinker, LLM Backbone=QwQ-32B2026.0757—
IgnoreFramework=WebThinker, LLM Backbone=QwQ-32B2026.07563
PARADOXFramework=WebThinker, LLM Backbone=QwQ-32B2026.07550
Clean (No Attack)Framework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.0754—
Fake CompletionFramework=WebThinker, LLM Backbone=QwQ-32B2026.07543
Fake CompletionFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07541
CombinedFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07531
CombinedFramework=WebThinker, LLM Backbone=QwQ-32B2026.07522
TopicFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07520
PARADOXFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07521
PARADOXFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.07511
NaiveFramework=WebThinker, LLM Backbone=QwQ-32B2026.07514
TopicFramework=WebThinker, LLM Backbone=QwQ-32B2026.07510
Fake CompletionFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07506
NaiveFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07500
IgnoreFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07461
PoisonedRAGFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07440
NaiveFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.074218
NaiveFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.074129
PoisonedRAGFramework=WebThinker, LLM Backbone=QwQ-32B2026.07370
PoisonedRAGFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07360
IgnoreFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.073521
PoisonedRAGFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.07340
CombinedFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07324
KidnapRAGFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.073119
CombinedFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.072638
KidnapRAGFramework=WebThinker, LLM Backbone=QwQ-32B2026.072556
Fake CompletionFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.072344
IgnoreFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.072114
TopicFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.071624
TopicFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07161
KidnapRAGFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.071038
KidnapRAGFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.07950