Pengyu Nie

[Liliana Hotsko @liliana_hotsko]

How do you give a code LLM knowledge of an entire repository without paying for it at every single query?

We introduce Code2LoRA: a hypernetwork that turns a repository into its own LoRA adapter. Repo knowledge baked into weights → zero inference-time token overhead.

[Yuntian Deng @ ICML @yuntiandeng]

I built a 3D character you can control with language instead of predefined buttons.

How? I compiled a neural program that turns language instructions into movements, using ProgramAsWeights.

Just type: "act excited, wave, dance, then sit proudly"

Try it: https://t.co/wnQcwbDJ3M

[Yuntian Deng @ ICML @yuntiandeng]

🚀 Launching ProgramAsWeights (PAW)!

Define functions in English → PAW compiles them into tiny neural programs → Run locally like normal Python functions.

A neural program combines discrete text + continuous LoRA to adapt a fixed small interpreter.

🔗 https://t.co/N6ISkMYP3P

[Larry Yinxi Li @LarryYinxiLi]

📌 Modern code LLMs rely on subword tokenization (like BPE). But these tokenizers have no idea about code grammar, they just merge characters by frequency.

So what happens when subword tokenization ignores grammar? 🤔

🤗 Our Paper : https://t.co/moUEFnXRGN

#LLM4Code
(1/n)

[Jiyang Zhang @zhangjiyang6]

We present exLong, a reasoning augmented large language model, for generating exceptional behavior tests, i.e., tests that check that code under test catches unwanted events and throws correct exceptions. exLong outperforms LLMs like GPT-3.5 and analysis-based tools. [1/5]🧵