Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
De-Linearizing Agent Traces: Bayesian Inference of Latent Partial Orders for Efficient Execution
Dongqing li, Zheqiao Cheng, Geoff Nicholls, Quyu Kong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use BPOP instead of Heuristics Miner | trace-only baselines · SRASearch | Edge F1: 0.43 -> 0.91 | Improved |
| use BPOP instead of Heuristics Miner | trace-only baselines · Epigenomics | Edge F1: 0.38 -> 0.79 | Improved |
| use BPOP instead of Heuristics Miner on cloud scenarios | process mining · Cloud-IaC-6 (IP-Cov=1.0) | Edge F1: 0.60 -> 0.95 | Improved |
| use BPOP instead of Majority / Heuristics Miner | graph recovery · Epigenomics | Feasibility: 0.00 -> 0.84 | Improved |
| use Inductive Miner instead of BPOP | BPOP · Cloud-IaC-6 | Edge F1 vs feasibility: Feasibility 0.85 -> 1.00; Edge F1 0.95 -> 0.46 | Mixed |
| increase trace diversity IP-Cov from 0.8 to 0.9 (ablation) | BPOP · Cloud-IaC-6 | Edge F1: 0.413 -> 0.872 | Improved |
| increase trace diversity IP-Cov from 0.8 to 0.9 (ablation) | BPOP compiled executor · Cloud-IaC-6 | Tokens per task: 17,479 -> ~583 | Improved |
| increase trace diversity IP-Cov from 0.6 to 0.9 (ablation) | BPOP compiled executor · Cloud-IaC-6 | Fallback rate (%): 50.0 -> 0.0 | Improved |
| vary noise epsilon over 0.005 to 0.05 (ablation) | BPOP · Cloud-IaC-6 (IP-Cov>=0.9) | Edge F1: varies by less than 0.02 | No clear effect |
| add LLM fallback to compiled execution (Hybrid vs Expert) (ablation) | BPOP executor · Cloud-IaC-6 six scenarios | Success rate (%): 66.7% -> 100.0% | Improved |
| compiled Expert execution instead of LLM Explore (CoT) | agent execution · Cloud-IaC-6 six scenarios | Total LLM tokens: 381,794 -> 3,501 | Improved |
| use Bayesian Queue-Jump likelihood instead of frontier-softmax | BPOP · WFCommons SRASearch | runtime: predicted runtime exceeding 1,000 hours | Hurt |
Closest papers by meaning
- Automata from Agent Traces: Failure and Next-Step Prediction (NeurIPS 2026 · similarity 0.85)
- Workspace-Bench 1.0: Benchmarking AI Agents on Workspace Tasks with Large-Scale File Dependencies (NeurIPS 2026 · similarity 0.85)
- Optimize Once, Execute Fast: Latency-Aware Multi-Agent Workflow Learning for Recurrent Queries (NeurIPS 2026 · similarity 0.85)
- Towards Direct Latent-Space Synthesis for Parallel Branches in LLM-Agent Workflows (NeurIPS 2026 · similarity 0.84)
- AgentHijack: Benchmarking Computer Use Agent Robustness to Common Environment Corruptions (ICML 2026 · similarity 0.84)
- Benchmarking Agentic Workflow Generation (ICLR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
