1. 01

    LLMs are memorizing, not generalizing enough

    They lean on memorized patterns instead of transferable, compositional structure — how do we train for genuine generalization?

  2. 02

    LLMs have no fast loop to verify against the real world

    Learning happens in slow, offline training runs. Without a high-frequency loop that checks claims against live, real-world facts, a model cannot ground or correct a general world model.

Research → a VR assistant

An interactive map from a standalone headset’s constraints to the papers that address them.

recent papers
  1. 012026safetypost-trainingin submission · ACM IUI 2027

    Do Not Be Angry at Your Agent: Insights from Human-AI Collaborative Problem Solving

    When an agent gets it wrong over a long multi-turn trajectory, should the human restart it, steer it neutrally, or steer it emotionally? Across twelve models and five benchmarks the best strategy depends on the agent — restarting wins for instruction-tuned models, emotional steering pays for reasoning-specialised ones — but one constraint survives: anger is never the right choice. A Jacobian-lens readout shows why, angry feedback pulls the deep-layer workspace toward apology and agreement and away from task content, turning a correction into sycophantic capitulation.

  2. 022026safetysynthetic datain submission · ARR 2026 August

    Should We Type or Talk to LLM Agents? A Comprehensive Study of Voice and Keyboard Input Perturbations

    Typing and speaking leave different signatures on a prompt — orthographic noise from keyboards, disfluency and restructuring from transcription. HIVE perturbs both channels to measure the cost. Voice perturbations lower accuracy on every instruction-tuned model tested and it is the structure of the transcription, not its fillers, that carries the cost; keyboard noise costs far less. Both trace to one cause: how many of the question’s tokens survive. A thinking budget recovers the keyboard channel almost entirely but leaves speech untouched.

    view
  3. 032026memorypost-trainingin submission · NeurIPS 2026

    SHRED: Retain-Set-Free Unlearning via Self-Distillation with Logit Demotion

    Most LLM unlearning methods need a curated retain set to stop general utility collapsing. SHRED needs only the forget set: the highest-information tokens in each instance become forget positions, the rest stay as benign anchors, and one top-k KL self-distillation objective demotes memorized logits while preserving the distribution elsewhere. Pareto-optimal forget/utility trade-off on four benchmarks, robust to relearning and membership-inference attacks.

    view
  4. 042026safetypost-trainingin submission · EMNLP 2026

    Expert Personas Improve LLM Alignment but Damage Accuracy: Bootstrapping Intent-Based Persona Routing with PRISM

    Persona effectiveness is task-type dependent: expert prompts consistently improve alignment-dependent tasks (safety, preference) but reliably damage pretraining-dependent knowledge retrieval. PRISM teaches models when to invoke a persona via intent-based self-modeling, preserving accuracy while keeping alignment gains.

    view
  5. 052026latencymemoryin progress · aiming ICLR 2027

    AttendTwice: Long-Context Inference via Dynamic Chunk-Level KV-Cache Selection

    Two attention paths. A cheap low-rank path scores which chunks of the KV cache matter for a query; the exact path then attends only over the chunks it selected, so most of the cache is never loaded. Selecting contiguous chunks rather than scattered tokens is what makes the gather fast enough to pay for itself — per-token selection is exact but its scattered access pattern costs more than it saves.

  6. 062025synthetic datasafetyACM ICMI 2025

    Multimodal Synthetic Data Finetuning and Model Collapse

    Studies how vision-language models degrade when fine-tuned on AI-generated multimodal data. Characterizes the collapse dynamics specific to the multimodal regime and proposes mitigation strategies that preserve diversity across modalities.

    view
  7. 072024pretraininglatencypreprint

    Lateralization MLP: A Simple Brain-inspired Architecture for Diffusion

    A brain-inspired MLP architecture with hemispheric lateralization applied to diffusion models. Shows competitive sample quality at reduced parameter count, suggesting structured asymmetry as an inductive bias for generative modeling.

    view
  8. 082024pretraininglatencyin submission · TMLR

    Static Key Attention in Vision

    A more efficient attention variant for vision transformers that pre-computes a static key projection, reducing per-token compute while maintaining downstream task performance.

    view
full list on Google Scholar
academic service
  • Reviewer · NeurIPS 2024–2026
  • Reviewer · ICLR 2024–2025
  • Reviewer · ICML 2024–2025
  • TA · DSCI 552 Machine Learning for Data Science (2023, 2025)
  • TA · CSCI 576 Multimedia Systems Design (2022, 2024, 2025)
  • TA · CSCI 567 Machine Learning (2022, 2024)
  • TA · CSCI 566 Deep Learning and Its Applications (2024)
  • TA · CSCI 544 Applied Natural Language Processing (2023)
Zizhao Hu
loading