Research map / Large language models / Language, safety and interpretability
Beyond Multi-Token Prediction: Pretraining LLMs with Future Summaries
Divyat Mahajan, Sachin Goyal, Badr Youbi Idrissi, Mohammad Pezeshki, Ioannis Mitliagkas, David Lopez-Paz, Kartik Ahuja
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use learned reverse-LM future summary instead of next-token prediction only | FSP-RevLM · ARC-Easy (8B) | pass@1: 71.8% -> 76.6% | Improved |
| Use FSP-RevLM instead of NTP | FSP-RevLM · GSM8K (8B) | pass@16: 0.716 -> 0.705 | Hurt |
| Use FSP-RevLM instead of MTP | FSP-RevLM · MATH (8B) | pass@16: 0.309 -> 0.351 | Improved |
| Use FSP-RevLM instead of DeepSeek-MTP at small scale | FSP-RevLM · ARC and code tasks (3B) | pass@1/pass@16: HumanEval+ 0.348 -> 0.305 | Hurt |
| Predict random future tokens (MTP-Skip) instead of the immediate next token (ablation) | MTP · GSM8K (8B) | pass@16: 0.678 -> 0.598 (tau 32) | Hurt |
| Use bag-of-words future summary instead of immediate-token MTP target (ablation) | MTP · GSM8K (8B) | pass@16: 0.678 -> 0.714 (tau 100) | Improved |
| Lengthen the BoW window from 12 to 100 (ablation) | FSP-BoW · ARC-Challenge (8B) | pass@1: 0.562 -> 0.459 | Hurt |
| Use a bag-of-words summary on path-star graphs instead of MTP (ablation) | MTP · Path-star G(2,8) | exact path accuracy: 0.48 -> 1.00 | Improved |
| Add more auxiliary heads to MTP (ablation) | MTP · Path-star G(2,8) | exact path accuracy: 0.48 (4 heads: 0.48) | No clear effect |
| Use RevLM summary instead of BoW on sibling discovery with many components (ablation) | FSP-BoW · Sibling discovery | steps to convergence vs NTP | Improved |
| Remove tf-idf reweighting from FSP-BoW (ablation) | FSP-BoW tau 12 · GSM8K (3B) | pass@16: 0.419 -> 0.431 | Mixed |
| Use a shallower reverse-LM depth (depth 2) in FSP-RevLM (ablation) | FSP-RevLM · 3B tasks | pass@16 / pass@1: MATH 0.212 -> 0.217; ARC-Easy 0.277 -> 0.265 | Mixed |
Closest papers by meaning
- Better & Faster Large Language Models via Multi-token Prediction (ICML 2024 · similarity 0.90)
- L-MTP: Leap Multi-Token Prediction Beyond Adjacent Context for Large Language Models (NeurIPS 2025 · similarity 0.88)
- Towards Auto-Regressive Next-Token Prediction: In-context Learning Emerges from Generalization (ICLR 2025 · similarity 0.88)
- Beyond Next Token Prediction: Patch-Level Training for Large Language Models (ICLR 2025 · similarity 0.87)
- Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning (ICML 2025 · similarity 0.87)
- Efficient Training-Free Multi-Token Prediction via Embedding-Space Probing (ICML 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
