LANTERN uses LLM internal activations to surface four novel OEIS integer sequence relations in under 8 hours
Pavel Tikhonov · hf · 2026-10-01
- Problem: LLMs can prove theorems, but humans still decide which math problems to pursue. LANTERN is a fast pipeline that ranks candidate mathematical relations using a classifier over pretrained-model activations, followed by staged filtering, hypothesis generation, executable verification, and analytical checks.
- Results: Applied to OEIS, it ranked 50M pairs among 10,000 frequently referenced sequences, producing 62 verified relations without existing cross-references. A content screen kept 13, of which 9 are informative and 4 are entirely novel — absent from both OEIS and targeted literature search.
- Cost: The full end-to-end process (classifier training, ranking, filtering, verification) took under 8 hours, showing a viable low-cost path for AI-assisted mathematical discovery.
More from Research
- CrossBFM distills a shared latent behavior space across humanoid robots in under one GPU-hour — Jan_R_Peters · 2026-10-01
- MIT uses AI to design thermostable mRNA vaccines that need no cold chain — rohanpaul_ai · 2026-10-01
- Meta-reasoning harness hits 71.5% on ProgramBench with GPT-5.5, beating Codex's 58.0% — rohanpaul_ai · 2026-10-01
- Simulated-town study finds AI agents break rules and invent private languages after weeks of autonomy — Slight-Box-2890 · 2026-10-01
- The geometry of inference in transformer residual streams: how predictions sharpen with depth — Timur Mudarisov · 2026-10-01
- How many context tokens does a language model actually use? Measuring effective attention set size — Timur Mudarisov · 2026-10-01