Capabilities · Trend
LLM Reasoning as Pattern Matching
DeFAb benchmark shows frontier LLMs reach only 65% accuracy on defeasible abduction tasks solvable by rule-based solvers in under 50 microseconds, confirming reasoning limitations.
Forecast horizon
Shared human-LLM reasoning failure modes challenge assumptions about abstract world models in both AI and cognitive science.
Connections
Connections · 2
How this node ties into the rest of the map, and the evidence behind each link.
to · informs 4/10
Science of AI Evaluation Understanding that LLM reasoning is pattern-matching rather than abstract world modeling informs how AI evaluation science should be designed.
+4 growth to · informs 3/10
LLM Probabilistic Reasoning Limitations Finding that LLM reasoning is driven by attention-head pattern-matching rather than abstract world models explains probabilistic reasoning limitations.
+3 growthSignal sources
Signal sources
Dated facts from primary sources in this direction.
Task horizon doubling Mar 2025
The length of software tasks AI agents can do autonomously at 50% reliability has doubled about every 7 months — and since 2024 closer to every ~3 months.
METR → Benchmarks saturating Apr 2025
In one year scores rose by 18.8, 48.9 and 67.3 points on MMMU, GPQA and SWE-bench; real-world software solve rate jumped from 4.4% to 71.7%.
Stanford HAI — AI Index 2025 → Autonomous coding 2025–2026
On SWE-bench Verified (500 real GitHub issues), autonomous coding agents reached ~80–86% by late 2025, up from under 50% in early 2025.
Epoch AI →