# LFM2 Technical Report

arXiv: https://arxiv.org/abs/2511.23404 · HTML: https://mlsanity.com/paper/2511.23404/

## Key facts

- Cited by 37 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 28 Nov 2025.

## Built on by (most cited first, 30 of 37)

- [Audio MultiChallenge: A Multi-Turn Evaluation of Spoken Dialogue Systems on Natural Human Interaction](https://mlsanity.com/paper/2512.14865/) (2025)
- [From Chatbot to Digital Colleague: The Paradigm Shift Toward Persistent Autonomous AI](https://mlsanity.com/paper/2606.14502/) (2026)
- [When Single Answer Is Not Enough: Rethinking Single-Step Retrosynthesis Benchmarks for LLMs](https://mlsanity.com/paper/2602.03554/) (2026)
- [Steerable Visual Representations](https://mlsanity.com/paper/2604.02327/) (2026)
- [MeMo: Memory as a Model](https://mlsanity.com/paper/2605.15156/) (2026)
- [Marco-MoE: Open Multilingual Mixture-of-Expert Language Models with Efficient Upcycling](https://mlsanity.com/paper/2604.25578/) (2026)
- [How Small Can 6G Reason? Scaling Tiny-to-Small Language Models for AI-Native Networks](https://mlsanity.com/paper/2603.02156/) (2026)
- [The Voice Behind the Words: Quantifying Intersectional Bias in SpeechLLMs](https://mlsanity.com/paper/2603.16941/) (2026)
- [6G Needs Agents: Toward Agentic AI-Native Networks for Autonomous Intelligence](https://mlsanity.com/paper/2605.01546/) (2026)
- [Joint Optimization of Tool Creation and Use for Large Language Model Agents](https://mlsanity.com/paper/2608.24571/) (2026)
- [LEMUR: Learned Multi-Vector Retrieval](https://mlsanity.com/paper/2601.21853/) (2026)
- [Directional Reasoning Trajectory Change (DRTC): Identifying Critical Trace Segments in Reasoning Models](https://mlsanity.com/paper/2602.15332/) (2026)
- [TiCo: Time-Controllable Spoken Dialogue Model](https://mlsanity.com/paper/2603.22267/) (2026)
- [Chronicle: A Multimodal Foundation Model for Joint Language and Time Series Understanding](https://mlsanity.com/paper/2605.20268/) (2026)
- [ATMA: Long-Context Language Modeling via Polar Attention and Gated-Delta Compression Memory](https://mlsanity.com/paper/2606.25156/) (2026)
- [Masked Diffusion Language Models are Strong and Steerable Text-Based World Models for Agentic RL](https://mlsanity.com/paper/2607.16204/) (2026)
- [Puro-2B: Poor Lab's Qwen2-1.5B Trained on RTX 5090 within 5090](https://mlsanity.com/paper/2608.27370/) (2026)
- [SafeLens: Deliberate and Efficient Video Guardrails with Fast-and-Slow Screening](https://mlsanity.com/paper/2605.17610/) (2026)
- [PluraMath: Extending Mathematical Reasoning Evaluation Beyond High-Resource Languages](https://mlsanity.com/paper/2607.05992/) (2026)
- [Multi-Agent Robotic Control with Onboard Vision-Language Models](https://mlsanity.com/paper/2607.07403/) (2026)
- [Let the Data Decide: Supervision Analysis, Capability Trade-offs, and Adaptive Objective Routing in Continued Pre-Training via Off-Policy Distillation](https://mlsanity.com/paper/2607.16246/) (2026)
- [Efficient Chain-of-Modality Reasoning via Progressive Compression for Spoken Language Models](https://mlsanity.com/paper/2607.19932/) (2026)
- [Summarize First, Download Later: Onboard VLMs for Bandwidth-Efficient Earth Observation](https://mlsanity.com/paper/2608.06959/) (2026)
- [Drive-to-Music: Context-Aware Generative Audio for In-Vehicle Experiences](https://mlsanity.com/paper/2608.12615/) (2026)
- [How Weight Encoding Affects Language Model Placement and Performance on the Apple Neural Engine](https://mlsanity.com/paper/2608.22110/) (2026)
- [Evaluating Context Segmentation in Locally Deployable SLMs for Cybersecurity CTF Tasks](https://mlsanity.com/paper/2609.12839/) (2026)
- [Continuous-Time Machine Learning: A Unified Mathematical Perspective](https://mlsanity.com/paper/2609.16710/) (2026)
- [A Multi-Modal Generative Model for Tomato Disease Leaves Understanding](https://mlsanity.com/paper/2609.19555/) (2026)
- [Minimally Invasive Steering of Language Models](https://mlsanity.com/paper/2609.30218/) (2026)
- [VisionPsy-Nano: Improving Accuracy, Efficiency, and Reliability in On-Device Vision-Language Models](https://mlsanity.com/paper/2609.31746/) (2026)

## Abstract (arXiv)

We present LFM2, a family of Liquid Foundation Models designed for efficient on-device deployment and strong task capabilities. Using hardware-in-the-loop architecture search under edge latency and memory constraints, we obtain a compact hybrid backbone that combines gated short convolutions with a small number of grouped query attention blocks, delivering up to 2x faster prefill and decode on CPUs compared to similarly sized models. The LFM2 family covers 350M-8.3B parameters, including dense models (350M, 700M, 1.2B, 2.6B) and a mixture-of-experts variant (8.3B total, 1.5B active), all with 32K context length. LFM2's training pipeline includes a tempered, decoupled Top-K knowledge distillation objective that avoids support mismatch; curriculum learning with difficulty-ordered data; and a three-stage post-training recipe of supervised fine-tuning, length-normalized preference optimization, and model merging. Pre-trained on 10-12T tokens, LFM2 models achieve strong results across diverse benchmarks; for example, LFM2-2.6B reaches 79.56% on IFEval and 82.41% on GSM8K. We further build multimodal and retrieval variants: LFM2-VL for vision-language tasks, LFM2-Audio for speech, and LFM2-ColBERT for retrieval. LFM2-VL supports tunable accuracy-latency tradeoffs via token-efficient visual processing, while LFM2-Audio separates audio input and output pathways to enable real-time speech-to-speech interaction competitive with models 3x larger. LFM2-ColBERT provides a low-latency encoder for queries and documents, enabling high-performance retrieval across multiple languages. All models are released with open weights and deployment packages for ExecuTorch, llama.cpp, and vLLM, making LFM2 a practical base for edge applications that need fast, memory-efficient inference and strong task capabilities.