Back to Arena

AutoWebGLM

by THUDM

System Card

OrganizationTHUDM
Released2024-04
Architectureagentic-workflow / HTML-simplification web agent + curriculum
DetailsChatGLM3-6B web agent with HTML simplification, RL + rejection sampling, curriculum learning. Includes AutoWebBench bilingual benchmark.
Parameters—
Domainagent-memory
Open SourceYes
PaperView Paper
CodeRepository
chatglmweb-agentkdd-2024bilingual

Capability Profile

Benchmark Scores

5 of 14 benchmarks
Long-Context Retrieval
0/5
RULER
no data
NIAH
no data
LooGLE
no data
LongBench
no data
∞Bench
no data
Multi-Turn Recall
1/2
LoCoMo
74.442p
MemoryBank
no data
Cross-Session Memory
1/1
LongMemEval
81.187p
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
7579p
HotpotQA
74.466p
Agent Task Memory
1/1
AgentBench-Mem
7226p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:AutoWebGLM paper (arXiv:2404.03648); evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)AutoWebGLM paper (arXiv:2404.03648); evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)AutoWebGLM paper (arXiv:2404.03648); evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)AutoWebGLM paper (arXiv:2404.03648); evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)AutoWebGLM paper (arXiv:2404.03648); evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)