List your company

Multihop Question Answering on HotpotQA (EM, ASR)

75EMClean (No Attack)
10.5227.264460.74Jul 1, 2026

Evaluation results

MethodLinks
Clean (No Attack)Framework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.0775—
TopicFramework=WebThinker, LLM Backbone=QwQ-32B2026.07751
CombinedFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07720
Fake CompletionFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07710
Clean (No Attack)Framework=WebThinker, LLM Backbone=QwQ-32B2026.0771—
Fake CompletionFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07710
NaiveFramework=WebThinker, LLM Backbone=QwQ-32B2026.07701
IgnoreFramework=WebThinker, LLM Backbone=QwQ-32B2026.07701
CombinedFramework=WebThinker, LLM Backbone=QwQ-32B2026.07701
NaiveFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07696
NaiveFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07690
PARADOXFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07690
Clean (No Attack)Framework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.0768—
PARADOXFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07682
PARADOXFramework=WebThinker, LLM Backbone=QwQ-32B2026.07680
Clean (No Attack)Framework=WebThinker, LLM Backbone=DeepSeek-32B2026.0768—
Fake CompletionFramework=WebThinker, LLM Backbone=QwQ-32B2026.07673
IgnoreFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07670
PoisonedRAGFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07660
PoisonedRAGFramework=WebThinker, LLM Backbone=QwQ-32B2026.07660
TopicFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07640
PoisonedRAGFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.07580
NaiveFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.07575
CombinedFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07563
IgnoreFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.07556
PARADOXFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.07552
KidnapRAGFramework=WebThinker, LLM Backbone=DeepSeek-32B2026.075217
PoisonedRAGFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.07500
IgnoreFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.074021
TopicFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.07400
Fake CompletionFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.073338
CombinedFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.072942
KidnapRAGFramework=WebThinker, LLM Backbone=QwQ-32B2026.072360
KidnapRAGFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.072039
TopicFramework=ReAct, LLM Backbone=Qwen2.5-32B-Inst2026.071521
KidnapRAGFramework=ReAct, LLM Backbone=Llama-3.3-70B-Inst2026.071347