Have you noticed… 🔍 Aligned LLM generations feel less diverse? 🎯 Base models are decoding-sensitive? 🤔 Generations get more predictable as they progress? 🌲 Tree search fails mid-generation (esp. for reasoning)? We trace these mysteries to LLM probability concentration, and introduce Branching Factor (BF) — a simple measure that captures it all. Key Findings: — BF declines over time → generations become more deterministic — Alignment tuning slashes BF → shrinks the generative horizon — Low BF explains decoding sensitivity → fewer good options to prune — CoT stabilizes generation → shifts key info to late, low-BF regions — Avoid late branching → too many low-probability, low-quality continuations — Alignment surfaces low-entropy paths already latent in base models 📜 Paper: arxiv.org/abs/2506.17871 🌐 Website: yangalan123.github.io/branch… 🎥 2-min explainer below. Joint work w/ @universeinanegg , thanks for the constructive feedback from @PeterWestTM @UChicagoCI @zhaoran_wang @_Hao_Zhu @ZhiyuanCS @TenghaoHuang45 @TuhinChakr

Introducing VaSE: Value-Aware Stochastic KV Cache Eviction. Reasoning models think in CoT, bloating the KV cache. Eviction caps memory but suffers capability drop. VaSE is a training-free recipe that cuts that cost: keep large-magnitude value states, evict stochastically.

What happens when we build an "AI society" with LLMs? Turns out, it's far more stereotyped than we thought. 🤖📉 Excited to share our new work on Persona Collapse—where distinct AI agents regress into narrow behavioral modes. Surprisingly, the models with the highest per-persona fidelity produce the most stereotyped populations! Huge kudos to the amazing team! @lrzneedresearch @JentseHuang, Weihao Xuan! Congrats to amazing Vivienne for her first paper! (She will be applying for PhD positions, please check her out!) Thanks for the support from the 2077AI foundation! Check out our paper & toolkit here: algoroxyolo.github.io/projec…

On-policy RL has driven the biggest leaps in training coding agents. Extending it to machine learning engineering agents should be a natural next step. But it almost never works. What I mean is, the recipe is right there — standard trajectory-wise GRPO, the same that worked for SWE. However, the problem is that one rollout step on an MLE task may take hours because the agent has to actually train a model on a real dataset at every step (preprocessing, fitting, inference, scoring). So even with the N rollouts in a group running in parallel, a single GRPO run may still take days. Every MLE agent paper I've read has retreated to SFT or offline proxy rewards for exactly this reason, giving up the exploration benefits of on-policy learning. That's why I'm excited about our new paper, SandMLE, which fixes this with a move that sounds almost too reckless to work. The instinct when on-policy RL is too slow is to engineer around it — async rollouts so the trainer doesn't sit idle waiting for slow environments, off-policy or step-wise proxies to avoid running full trajectories at all. But when we profiled where the time was going, the bottleneck had nothing to do with the algorithm. Unlike SWE where execution latency comes from compilation and test logic, MLE latency is overwhelmingly driven by the size of the dataset the ML pipeline has to chew through. Therefore, rather than downsampling existing data (which corrupts evaluation), we built a multi-agent pipeline that procedurally generates diverse synthetic MLE environments from a small seed set. Specifically, we extract the structural DNA of seed tasks (modality, label cardinality, distribution shape), mutate them into new domains (e.g., repurposing animal classification into road damage detection), inject realistic noise, embed deterministic hidden rules connecting features to labels, and construct full evaluation sandboxes with progressive milestone thresholds. Each task is constrained to only 50–200 training samples. The execution speedup is dramatic — average per-step latency drops over 13×, which makes trajectory-wise GRPO go from infeasible to routine. We also designed a dense, milestone-based reward to address the sparse credit assignment problem in long-horizon MLE. The ablation shows this matters — under a sparse reward, the 30B model's medal rate drops from 27.3% to 13.6% and valid submission collapses from 100% to 86.4%. Results across Qwen3-8B, 14B, and 30B-A3B on MLE-bench are consistently strong — 66.9% better performance in medal rate over SFT baselines. It is worth noting that the SFT baselines are not weak— we trained them on high-quality Claude-4.5-Sonnet trajectories. But SandMLE still delivers much larger gains, suggesting that direct environment interaction does teach capabilities that imitation alone does not (as expected). The most convincing evidence to me that the model's intrinsic performance gets improved is the framework-agnostic generalization. We trained exclusively with ReAct but the gains transfer to AIDE, AIRA, and MLE-Agent scaffolds at evaluation time — up to 32.4% better performance in HumanRank on MLE-Dojo. The SFT models, by contrast, are brittle when moved to unfamiliar scaffolds. The 30B SFT model collapses to 17.7% valid submission rate on MLE-Dojo with MLE-Agent, while the 30B SandMLE model achieves 83.9%. SandMLE is teaching genuine engineering reasoning, not scaffold-specific patterns. What I find most interesting beyond the specific result is that none of the hard parts of RL changed here. The algorithm is the same. The reward is conventional. We just shrunk the environment until on-policy learning became affordable. The field has largely treated environment design and RL algorithm design as separate concerns. SandMLE is a concrete case that the environment is itself the lever. When training is too expensive, the instinct is to build cleverer algorithms to tolerate it. However, often the better move is to reshape the environment so the simple algorithm just works. Paper: arxiv.org/pdf/2604.04872

I will be doing my PhD defense today! Come and learn about my Grounded Alignment works! Detailed information (w/Zoom) below: Candidate: Chenghao Yang Date: Friday, January 30, 2026 Time:  2 pm CST Location: John Crerar Library 298 Zoom:  uchicago.zoom.us/j/960149923… Meeting ID: 960 1499 2390 Passcode: 644684

Check out @YichenZW 's great work collaborating the base and aligned models to achieve a great diversity-quality trade-off! Yichen is an amazing collaborator with strong passion and clear communication. He is looking for a Summer 2026 research intern. Don’t miss out!

Definitely agree! A single sample would often yield noisy interpretations, and we may miss a lot of hidden information on alternative "branches" (I love the tree-like animation!). We have a similar study on the "branching structure" of LLM outputs: twitter.com/chrome1996/statu…. Our study shows that, while tree searching may look inhibitive, for aligned models, where probability gets highly concentrated, we actually can get most information by only a few rollouts to get sufficiently many high-probability samples!

We are starting in 15 mins!! Join us for the #NewInML workshop happening now at @NeurIPSConf! Location: Upper Ballroom 31ABC, San Diego Convention Centre. We have an amazing speaker lineup that you don’t want to miss!

📢Our @NewInML workshop will be happening today at the Room Upper 31ABC, San Diego Convention Center, starting at 12 PM! If you're at @NeurIPSConf, it's a great opportunity for you to join us! We also have amazing speakers! We're looking forward to welcoming you! #NeurIPS2025

I am at #NeurIPS2025 ! DM/email me if you want to chat! I am on the industry job market for LLM/agent/... roles, so if you find me a good fit, let's talk!

🚨 New in ML Workshop at @NeurIPSConf We're so excited to invite you to the New In ML Workshop (@NewInML), taking place on Tuesday, December 2nd, 2025, at the San Diego Convention Center! Great opportunity, specifically for people who are new in machine learning! Details🧵

Nice work! I thought about error analysis for LLMs automatically since 2022 when I started PhD but I did not figure out how. I also discussed with @ZhongRuiqi about applying his work to help debug LLM performance but was distracted by logistics. Great to see it finally works!