List your company
ResearchFormerly WizWand

Scientific Reasoning on ScienceWorld Unseen

62Average RewardCo-Evolving Agents
-1.76260814.79114631.344947.898654Nov 27, 2025Dec 28, 2025Jan 28, 2026Feb 28, 2026Mar 31, 2026May 1, 2026Jun 2, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Co-Evolving AgentsAdaptation=Fine-tuning, Backbone=Llama-2-7B-Chat2025.1162
Co-Evolving AgentsBackbone=Qwen3-4B-Instruct-25072025.1158.5
Llama-2-7B-Chat + ETOAdaptation=Fine-tuning, Backbone=Llama-2-7B-Chat2025.1155.5
ETOBackbone=Qwen3-4B-Instruct-25072025.1155.2
Llama-2-7B-Chat + RFTAdaptation=Fine-tuning, Backbone=Llama-2-7B-Chat2025.1154.3
Llama-2-7B-Chat + PPOAdaptation=Fine-tuning, Backbone=Llama-2-7B-Chat2025.1151.7
Llama-2-7B-Chat + SFTAdaptation=Fine-tuning, Backbone=Llama-2-7B-Chat2025.1141.9
SFTBackbone=Qwen3-4B-Instruct-25072025.1140.8
GPT-4Adaptation=In-context2025.1138.1
GPT-3.5-TurboAdaptation=In-context2025.1110.5
DELTAMEMLLM=DeepSeek-V4-flash, Decoding Temperature=0, Prompt Framework=ReAct, Node Embeddings=E5-base-v2, Learning Protocol=Online learning, in-context demonstration=nicl = 12026.060.8688
SynapseLLM=DeepSeek-V4-flash, Decoding Temperature=0, Prompt Framework=ReAct, Node Embeddings=E5-base-v2, Learning Protocol=Online learning, in-context demonstration=nicl = 12026.060.8558
AWMLLM=DeepSeek-V4-flash, Decoding Temperature=0, Prompt Framework=ReAct, Node Embeddings=E5-base-v2, Learning Protocol=Online learning, in-context demonstration=nicl = 12026.060.7453
No MemoryLLM=DeepSeek-V4-flash, Decoding Temperature=0, Prompt Framework=ReAct, Node Embeddings=E5-base-v2, Learning Protocol=Zero-shot, in-context demonstration=nicl = 12026.060.7186
RBankLLM=DeepSeek-V4-flash, Decoding Temperature=0, Prompt Framework=ReAct, Node Embeddings=E5-base-v2, Learning Protocol=Online learning, in-context demonstration=nicl = 12026.060.6898