Prompt EngineeringIman Mirzadeh et al. (Apple)Apple

Apple researchers present GSM-Symbolic

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

GSM-Symbolic presented — performance drops when templates swap the numbers and names, or when an irrelevant clause is added

Why it matters

Evidence that math reasoning may not be formal reasoning

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2410.05229

Added in Week 35, 2026 · See that week

On the map — ① Prompt Engineering line

  1. Previous stopChain-of-Verification
  2. This recordGSM-Symbolic
  3. Next stopGPT-4.1
See it on the map

Same week

Xi Jiang et al. release the MMAD benchmarkPaperAIST announces ABCI 3.0OfficialGENIAC selects 20 AI foundation model projects and 3 data-use demonstrationsOfficialMeta presents TorchTitanPaper

Nearby on the same axis

Anthropic launches prompt improver in the ConsoleOfficialZayne Sprague et al. publish "To CoT or not to CoT?"PaperMIPRO publishedPaperMuennighoff et al. present s1Paper