V2EX / 2026-11-10

[程序员] Harness 跑分比赛， FrontierHarness benchmark v1.1 直播

感觉还挺有意思的 FrontierHarness Eval v1.1 直播跑分，21 harness × 5 model × 62 task × 4 次尝试：总计 26,040 次运行，比此前的 v1.0 360 次的有大幅提升（规模扩大了 72 倍） https://x.com/guanlan/status/2120192585235046712

Read original source

Source page screenshot: [程序员] Harness 跑分比赛， FrontierHarness benchmark v1.1 直播 — Source page screenshot · www.v2ex.com2026-11-10 UTC

Source record

Source: V2EX
Published: 2026-11-10 16:53:01 UTC
First discovered: 2026-11-10 17:30:12.471Z
Last collected: 2026-11-10 17:30:12.471Z
Author: sasuke688848
Topic: AI infrastructure
Classification: Models & research
Evidence level: E1 · source record; conclusions unverified

Research analysis

Rule-based research guidance derived from the source record. Audience and commercial demand require validation.

Research claim

Source account: 感觉还挺有意思的 FrontierHarness Eval v1.1 直播跑分，21 harness × 5 model × 62 task × 4 次尝试：总计 26,040 次运行，比此前的 v1.0 360 次的有大幅提升（规模扩大了 72 倍） https://x.com/guanlan/status/2120192585235046712

Evaluation record

Recent records to compare: “Next-Concept Modeling: Fast, Parallel Coarse-to-Fine Text Generation” (Hacker News, Nov 9, 2026); “Forcing a Letter Suppresses Refusal: Mechanistic Evidence from WMDP” (arXiv, Nov 9, 2026). Shared themes are navigation hints; they have not been verified as the same need or independent corroboration.

What the result supports

Interpret results only for the reported model version, task, data and inference budget. Community votes, downloads or a benchmark position do not establish reliability in a different workflow.

Validity gaps

Check sample selection, baseline configuration, uncertainty, contamination risk, negative results and access to code or weights. No reproduction by WeWorkBuddy is recorded here.

Verification plan

Read the paper or model card and its evaluation appendix. Reproduce one bounded result against the stated baseline, then test a representative user task and record where the conclusion fails.