Verdict
Evenly split 2–2 across 4 shared benchmarks.
Llama 2-70B 2 · PaLM 2-L 2 · higher isn't always better — check the caveats below.
Reported scores — protocols may differ. How we compare
Commonsense Reasoning: PaLM 2-L leads 2–0 across 2 shared commonsense reasoning benchmarks (largest gap: Winogrande, 66 vs 60.4 — single evaluator).
Evenly split 2–2 across 4 shared benchmarks. Llama 2-70B leads 2 benchmarks and PaLM 2-L leads 2. Higher isn't always better — see the integrity caveats on each benchmark.
Llama 2-70B and PaLM 2-L have 4 benchmarks in common in our data — those are the only rows where a direct, apples-to-apples comparison is drawn.
PaLM 2-L leads commonsense reasoning 2–0.