Skip to content
Papers.

When EOS Tokens Disagree: Understanding Length Inflation in On-Policy Distillation research paper by Microsoft, 2026

Microsoft · Sep 17, 2026 · Foundation models · 0 citations · 107 upvotes · unverified

Read on arXiv

What it shows

We study length inflation in on-policy distillation (OPD), where student responses can become excessively long and even exhaust the generation budget.

UnverifiedHugging Face's summary; not yet checked by hand.

More from Microsoft

All 45
PaperCitations
The Tasteful Agent: Measuring and Improving Taste in Long-Horizon TasksLLM agents increasingly work on long-horizon tasks, and the decisions they make along the way, such as which hypothesis to test or which implementation to build on, determine the outcome of the whole run.Agents and evaluation · Sep 2026 · Unverified0
When2Think: Learning Difficulty-Aware Length Control for Efficient Hybrid Reasoning ModelsLarge Reasoning Models (LRMs) achieve strong performance on complex tasks but exhibit systematic inefficiency: they often overthink easy problems and underthink hard ones.Reasoning · Sep 2026 · Unverified0
BI-Agent and BI-Bench: Towards Automating End-to-End Business IntelligenceBusiness intelligence (BI) is a cornerstone of enterprise decision-making and is widely used by enterprise users in software such as Power BI and Tableau.Agents and evaluation · Sep 2026 · Unverified0
StudentSim: Training LLM-based Student SimulatorsStudentSim trains per-student simulators that answer like a given learner and change their answers under a tutor's guidance.Applied AI · Sep 20260
AutoSaddler: Automatic Harness Optimization with Durable Updates from Agent Execution TracesAutoSaddler automatically improves LLM agent harnesses via offline failure-driven optimization, boosting performance on long-horizon benchmarks.Agents and evaluation · Aug 2026 · Unverified2
Agent Lightning v1.0: Towards Harnessed Agentic RLAgent Lightning v1.0 enables reproducible reinforcement learning for arbitrary agent harnesses, substantially improving coding-agent performance with minimal data and compute.Agents and evaluation · Aug 2026 · Unverified2
OasisKV: Scaling In-Decode KV Cache Beyond HBM with Lookahead Sparse PrefetchingOasisKV improves LLM inference throughput by storing full KV caches in lower memory tiers and prefetching only relevant entries into HBM using speculative-decoding lookahead predictions.Inference and efficiency · Aug 2026 · Unverified0
Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation ModelMage-VL is a codec-native streaming vision-language model that uses motion-aware tokenization and a dual-system architecture to enable efficient real-time multimodal perception and reasoning.Inference and efficiency · Jul 2026 · Unverified3

Other foundation models papers

Topic
PaperCitations
BERT: Pre-training of Deep Bidirectional Transformers for Language UnderstandingPre-training a Transformer to fill in masked words, then fine-tuning it, beat task-specific models across language understanding tests.Google · Oct 2018121k
Language Models are Few-Shot LearnersGPT-3, a 175B-parameter model, does new tasks from a few examples in the prompt, with no fine-tuning.OpenAI · May 202063.7k
GPT-4 Technical ReportA multimodal model that passes a simulated bar exam with a score around the top 10% of test takers.OpenAI · Mar 202327.3k
LLaMA: Open and Efficient Foundation Language ModelsOpen models from 7B to 65B trained only on public data; the 13B model beats GPT-3 on most benchmarks.Meta · Feb 202321.7k
The Llama 3 Herd of ModelsOpen models up to 405B parameters that match leading closed models on many tasks.Meta · Jul 202418.9k
Qwen3 Technical ReportQwen3, a unified series of large language models, integrates thinking and non-thinking modes, reduces computational resources, and achieves state-of-the-art performance across various tasks and languages.Alibaba (Qwen) · May 2025 · Unverified8,100
About this paper
Authors
Yuxiao Yang, Tianrun Yu, Shangzhe Li and 6 more
arXiv
2609.20511 · PDF
Citations
0, 0 influential · Semantic Scholar
Upvotes
107 · Hugging Face
Code
github.com/UNCSciML/opd-eos
Lab
Microsoft · on Companies · on Acquisitions · on Quarterly · on Paydays · on Releases · on TechConf

Changes

What changed
Influential citationsfirst count: 0Sep 25, 2026
Citationsfirst count: 0Sep 25, 2026
New paperFound by the weekly scan, unverifiedSep 25, 2026

Sources: each lab's own papers and arXiv, with citation and upvote counts from Semantic Scholar and Hugging Face. One-line summaries are for orientation, not a substitute for the paper. Logos via logo.dev; trademarks belong to their owners.

New papers by email

Monday afternoons, only in weeks with new papers from the labs.

Double opt-in. Unsubscribe any time.