←
BangLiu76 posts
BangLiu avatar

BangLiu

@BangL93

I am an associate professor at University of Montreal and Mila.

◷ Joined June 2023▧ 18 Photos and videos
BangLiu avatar
BangLiu@BangL93·Oct 4

Across four benchmarks, the full Qwen3.5-27B configuration achieves a 10.1% relative improvement in average accuracy over the strongest baseline, MemPro, while using approximately 71% fewer context tokens and delivering 2.1× faster inference.

BangLiu avatar
BangLiu@BangL93·Oct 4

One key design: higher-level memories serve as indexes into the original evidence. At query time, the system navigates the hierarchy as needed, stopping once sufficient evidence is found while preserving traceability to the underlying records.

BangLiu avatar
BangLiu@BangL93·Oct 4

Through program evolution, it rewrites memory-management code to improve what to store, how to retrieve it, and when to dig deeper, using evaluation to select effective changes.

BangLiu avatar
BangLiu@BangL93·Oct 4

Looking up a detail, connecting evidence across conversations, and reusing a skill each place different demands on how information is organized and retrieved. MemCodex organizes raw records, summaries, structured knowledge, and reusable skills into a memory hierarchy

BangLiu avatar
BangLiu@BangL93·Oct 4

Our new work, MemCodex, takes an initial step toward this vision—and goes further: enabling agents to design their own memory systems. Different tasks call for different ways of remembering.

BangLiu avatar
BangLiu@BangL93·Oct 4

🧠 An agent’s long-term memory can be programmed and improved by the agent itself. Over a year ago, we began exploring whether the layering principles behind computer networks’ OSI model could inspire a memory framework for agents, with clearly defined layers and shared interface

Post attachmentPost attachment
BangLiu avatar
BangLiu@BangL93·Sep 23

These results do not establish a quantum computational advantage. They give us a concrete starting point for exploring larger quantum foundation models—and understanding what quantum computation can contribute to learning and generation.

BangLiu avatar
BangLiu@BangL93·Sep 23

With support from IVADO–PINQ², we also ran the 16–64-qubit circuits on IBM’s 156-qubit processor in Québec, retaining most of the model’s performance under real hardware noise on selected benchmark subsets.

BangLiu avatar
BangLiu@BangL93·Sep 23

Using just one-tenth of the fine-tuning data used by the classical baselines, the 64-qubit version outperformed LoRA on the same backbone by 3.67 percentage points, averaged across six benchmarks.

BangLiu avatar
BangLiu@BangL93·Sep 23

HyperQ is a lightweight circuit hypernetwork generates circuit parameters dynamically from each token’s context. Across the three circuit sizes we tested—16, 32 and 64 qubits—performance improved as the circuits grew wider.

BangLiu avatar
BangLiu@BangL93·Sep 23

The “quantum brain” hypothesis sparked our curiosity: what could quantum computing bring to large language models? After more than a year of research, we’re sharing HyperQ, which integrates trainable quantum circuits into a 1.1-billion-parameter diffusion language model.

Post attachmentPost attachmentPost attachment
BangLiu avatar
BangLiu@BangL93·Jun 25

I gave a talk on ``Beyond Capability: Toward agents we can trust, afford, and live with'' for Duke-NUS-Chen Institute Boot Camp for AI & Applications. Slides available on my homepage.

Post attachmentPost attachmentPost attachmentPost attachment