Skip to content
@nokia-applied-research

nokia-applied-research

README.md

Nokia Applied Research

We build production agents — systems that have to run, retrieve, and reason under real constraints.

Liang Wu  Liang Wu  ·  Email address


Open source

AnyJev GitHub stars
Turn any LLM into a typed decision model with real probabilities. No training.
pip install anyjev

Agent efficiency
WiSP — expert paging for MoE serving on vLLM
SPECTRA — KV cache compression past the 2-bit cliff
Trace — reliable context compression for long-horizon agents

Latest

  • Sep 2026   PAIR on arXiv: adapting context compression for long-horizon agents with counterfactual continuations.
  • Sep 2026   AnyJev v0.2.0: every decision level, L2 included, now serves on vLLM.
  • 2026   Papers at KDD, CIKM, ICDM, EMNLP and COLM.

Research

Agent efficiency

PAIR — Adapting Context Compression for Long-Horizon Agents with Counterfactual Continuations
Guanghui Min, Liang Wu, Mingjia Shi, Yinhan He, Mayank Darbari, Liangjie Hong, Chen Chen
arXiv 2026  ·  paper

SPECTRA — Pushing the KV Cache Beyond the 2-Bit Cliff via Spectral Transform Coding
Jiamu Zhang, Liang Wu, Kelly Wan, Hanjie Chen, Liangjie Hong
arXiv 2026  ·  paper  ·  code

TRACE — Toward Reliable Context Compression for Long-Horizon Agents: An Empirical Study of Execution Instability
Guanghui Min, Liang Wu, Mayank Darbari, Chen Chen, Liangjie Hong
arXiv 2026  ·  paper  ·  code

WiSP — A Working-Set View of Mixture-of-Experts Serving on Extremely Low-Resource Hardware
Jiamu Zhang, Liang Wu, Mayank Darbari, Liangjie Hong
arXiv 2026  ·  paper  ·  code

Domain LLMs

Telecom and networks

Towards Resilient and Autonomous Networks: A BlueSky Vision on AI-Native 6G
Liang Wu, Kelly Wan, Mayank Darbari, Liangjie Hong
KDD 2026  ·  paper

KAC — KPI-Aware Multimodal Anomaly Detection for 5G and Open RAN Telemetry
Chiman Salavati, Liang Wu, Kelly Wan, Mayank Darbari, Liangjie Hong
ICDM 2026  ·  code

TelecomAudit — Origin-Aware Benchmark Auditing and Calibration for 5G Anomaly Detection
Chiman Salavati, Liang Wu, Kelly Wan, Mayank Darbari, Liangjie Hong
CIKM 2026  ·  code

Mining the Self-Rewriting System: A Vision for Data Mining on AI-Native Infrastructure
Mayank Darbari, Chiman Salavati, Liang Wu, Dawei Wang, Jiong Zhang, Kelly Wan, Krishna Chakka, Liangjie Hong
ICDM 2026

Recommendation and search

PageLLM — A Multi-Grained Reward Framework for Whole-Page Optimization with Large Language Models
Xinyuan Wang, Liang Wu, Dongjie Wang, Yanjie Fu
CIKM 2026  ·  paper

SAPO — Step-Aligned Policy Optimization for Reasoning-Based Generative Recommendation
Zaiyi Zheng, Liang Wu, Guanghui Min, Yaochen Zhu, Liangjie Hong, Chen Chen, Jundong Li
arXiv 2026  ·  paper

Vision and time series

Self-Evolving Code-with-Image Reasoning
Tianze Yang, Liang Wu, Ruitong Sun, Yucheng Shi, Yanqiao Wang, Mayank Darbari, Ninghao Liu, Jin Sun, Liangjie Hong
arXiv 2026  ·  paper  ·  data

LLM Agents for Time-Series: A Survey
Yilong Chen, Xiao Qin, Chenghao Liu, Liang Wu, Noelle I. Samia, Kaize Ding
EMNLP 2026  ·  paper

Skill optimization

Field Aware Agent Skill Retrieval
Paimon Goulart, Liang Wu, Kelly Wan, Evangelos E. Papalexakis, Liangjie Hong
COLM 2026  ·  paper

Popular repositories Loading

  1. AnyJev AnyJev Public

    Turn any LLM into a Jev-style decision model: typed decisions, real probabilities, no training. (continue updating, welcome any issue and PR request)

    Python 1k 131

  2. WiSP WiSP Public

    Routing-aware expert paging for memory-constrained MoE serving — a vLLM plug-in with byte-identical outputs and joint expert/KV memory allocation (arXiv:2606.21868)

    Python 12 3

  3. Trace Trace Public

    Python 6 1

  4. SPECTRA SPECTRA Public

    Transform coding for the LLM KV cache — training-free spectral latent + reverse water-filling bit allocation, pushing past the 2-bit cliff (arXiv:2608.07915)

    Python 3 2

  5. .github .github Public

    Organization profile README

  6. Gated-Context-Memory-Pub Gated-Context-Memory-Pub Public

    Python

Repositories

Showing 7 of 7 repositories
  • AnyJev Public

    Turn any LLM into a Jev-style decision model: typed decisions, real probabilities, no training. (continue updating, welcome any issue and PR request)

    nokia-applied-research/AnyJev's past year of commit activity
    Python 1,012 Apache-2.0 131 7 (2 issues need help) 5 Updated Oct 2, 2026
  • Scaling_down_law Public Forked from XueqiC/ICLR27_scaling_down

    The code repo for paper "Capability Scaling-Down Laws for LLM Compression".

    nokia-applied-research/Scaling_down_law's past year of commit activity
    Python 0 2 0 0 Updated Oct 1, 2026
  • .github Public

    Organization profile README

    nokia-applied-research/.github's past year of commit activity
    0 0 0 0 Updated Oct 1, 2026
  • WiSP Public

    Routing-aware expert paging for memory-constrained MoE serving — a vLLM plug-in with byte-identical outputs and joint expert/KV memory allocation (arXiv:2606.21868)

    nokia-applied-research/WiSP's past year of commit activity
    Python 12 3 0 0 Updated Sep 15, 2026
  • Gated-Context-Memory-Pub Public
    nokia-applied-research/Gated-Context-Memory-Pub's past year of commit activity
    Python 0 0 0 0 Updated Sep 7, 2026
  • SPECTRA Public

    Transform coding for the LLM KV cache — training-free spectral latent + reverse water-filling bit allocation, pushing past the 2-bit cliff (arXiv:2608.07915)

    nokia-applied-research/SPECTRA's past year of commit activity
    Python 3 MIT 2 0 0 Updated Aug 12, 2026
  • Trace Public
    nokia-applied-research/Trace's past year of commit activity
    Python 6 1 0 0 Updated Aug 6, 2026

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Python