Changlong Yu

Applied Scientist @Amazon | LLM alignment and MLSys | Prev. @HKUST @MSFTResearch

[Zoey Sha Li @ZoeyLi20]

We observed a similar phenomenon with our own runs: when extrapolated to a horizon of thousands of steps, the early direction found by RELEX overshoots and performanace degrades.

[Zheye Deng @dzy97]

🏅 GPT-OSS-120B + Arbiter reaches gold at IOI 2026

Introducing Arbiter, a code verifier agent. Arbiter + GPT-OSS-120B scores 385.46/600, past the 361.12 gold cutoff.

We achieve gold-level performance at <100 generations per subtask, whereas previous search-based systems draw 5k. We're not just scaling TTC, but scaling it efficiently through boosting sequential self-correction.

📝 Project page: https://t.co/X33uUAz8Gb

#IOI2026
🧵👇

[Changlong Yu @yucl08]

Harness evolution from powerful proposers might gain large improvements from task overfitting (shortcut) or more test time compute. Careful attribution and new benchmarks for measuring generalization are indeed necessary.

[Wenxuan Ding @Wenxuan_Ding_]

Harness evolution brings a 0.8B model to 100% on ALFWorld. But how much is generalizable improvement vs. overfitting or test-time scaling?

Our method, Harness Delta Attribution, analyzes gains and attributes them to these categories. On 4 benchmarks, lots of the gain is overfitting or due to higher compute usage. Future evaluation should study this more!

Post: https://t.co/QnzjknULFH

[Wenxuan Ding @Wenxuan_Ding_]

Harness evolution brings a 0.8B model to 100% on ALFWorld. But how much is generalizable improvement vs. overfitting or test-time scaling?

Our method, Harness Delta Attribution, analyzes gains and attributes them to these categories. On 4 benchmarks, lots of the gain is overfitting or due to higher compute usage. Future evaluation should study this more!

Post: https://t.co/QnzjknULFH

[OphilusAI @OphilusAI]

Introducing Khora, our first multiplayer world model.

The demo is LIVE. Join an 8-player real-time deathmatch in one shared world now!

Built by Ophilus in collaboration with @RhOS_AI , Khora is designed to scale toward Interactive Worlds with unlimited players, with more generated maps and richer gameplay interactions to come.

Special thanks to @alibaba_cloud and FITCLOUD for infra and MaaS support, and to @DeemosTech for technical input.

[Zhenghao Xu @ZhenghaoXu0]

@Kimi_Moonshot used policy mirror descent (PMD) for RL in Kimi k1.5/k2. Most take it simply as PG+KL with an updating anchor, but this is not the full story. Check our blog for some interesting findings about this algorithm: https://t.co/6PgwAcjgp5

[Liang Qiu @liangqiu_1994]

Something not studied here: the joy human researchers get from working on a problem purely out of curiosity, and from collaborating with others.

[Yuwei Zhang @YuweiZh49446108]

On-policy self-distillation is a promising direction for learning from rich textual feedback. But can it really learn from failed trajectories?

Our answer: not quite -- unless we let the model actively interpret them.

🧵1/N