Perplexity open-sources WANDR: 500 research tasks, 170,495 evidence points, URL-rechecking grader
thisguyknowsai · x · 2026-11-02
Perplexity open-sourced WANDR (Apache-2.0, with an arXiv technical report), the benchmark behind the research engine in Perplexity Computer, built from de-identified real user cases.
- 500 realistic research tasks spanning 170,495 required evidence points: competitive research, due diligence, literature reviews, market analysis, product comparisons, talent sourcing
- "Wide" tasks find every qualifying company/paper/person; "Deep" tasks back every claim with evidence
- No fixed answer key: the grader re-fetches every URL the agent cited and verifies the evidence actually supports each record
- Repo is split into four layers: editable source data, a deterministic Harbor adapter, generated task packages, and a generic Relay agent; reproducible with Python 3.12, Docker and provider API keys
More from coding & agent
- InfoQ: A Systemic Approach to Memory, Knowledge, and Context in Agentic AI Architectures — AdiPolak · 2026-11-02
- Autonomous's Harness Pro uses an ESP32-P4, not a mini PC, to monitor AI coding agents — dee_hw · 2026-11-02
- RL-trained LFM2.5-24B adapter matches accuracy with 20% fewer tool calls — yb2698 · 2026-11-02
- Debugging production AI agents: error surfaces at step 5, root cause at step 3 — Ok-Machine7983 · 2026-11-02
- RAGFlow 1.0 preview: from RAG engine to agent-native database — newpeak · 2026-11-02
- Kitsune: open-source MCP server that shares your Firefox with a coding agent — Sad-Interview9085 · 2026-11-02