Amelia Dai

phd student @UChicago, previously NYU

[Amelia Dai @ameliadai_]

@JunyeobB @linapiuc @haifengxu0 @_elsehow @nadja_flechner @EzraKarger thank you!!

[Amelia Dai @ameliadai_]

CruxBench is accepted to #NeurIPS2026! 🎉

We evaluate not the answers, but the questions worth asking: can LLMs discover the cruxes that move beliefs about the future?

🔗 https://t.co/sjruGnbSxe

with @linapiuc, @haifengxu0, and FRI team: @_elsehow, @nadja_flechner, @EzraKarger

[Amelia Dai @ameliadai_]

Many thanks to my co-first author @linapiuc, my supervisor Prof. @haifengxu0, and our amazing collaborators at @Research_FRI for all the work, ideas, and support🙏

[Amelia Dai @ameliadai_]

(5/5)
Details matter. Models often pick similar events, but small differences in thresholds, deadlines, or resolution criteria can decide whether a crux catches the market move.

[Outerloop @outerloop_sci]

Outerloop 0.1.0 is out. A fleet of AI agents that improve your benchmarks on your code. Agents propose changes, run them on your cluster, and open pull requests when the metrics improve. We are open source. Let's embrace open science! Get started at https://t.co/tkGdjrUR9F

[Amelia Dai @ameliadai_]

Check out our work at #ICML2026! Will be presenting Beta Bernoulli Calibrator and our live forecast agent at the AI forecast workshop @ProphetArena on Jul 11th. Come say hi😃

[agentic learning ai lab @agentic_ai_lab]

Agentic Learning AI Lab @agentic_ai_lab presents at #ICML2026 in Seoul✈️🇰🇷 Come check out our latest research!

📍Main Conference Papers
Temporal Straightening for Latent Planning
Tue Jul 7 10:30am Hall A #1509

Context Tuning for In-Context Optimization
Wed Jul 8 5pm Hall A #2510

📍Workshop Papers
The Surprising Effectiveness of Deleting Weights in LLM Reasoning and Adaptation
Fri Jul 10 Rm 318 | Foundations of Deep Generative Models Workshop

Seeking the Unfamiliar but Memorable: Conceptual Creativity as Meta-Learning
Sat Jul 11 Rm 401 | Human-AI Co-Creativity Workshop

Aligning LLMs with Human Uncertainty: A Beta-Bernoulli Calibrator for LLM Forecasting
Sat Jul 11 Gr. Ballroom 104–105 | Forecasting as a New Frontier of Intelligence Workshop

Alive and Predicting: A Live Evaluation of Multi-Step Forecasting Agents
Sat Jul 11 Gr. Ballroom 104–105 | Forecasting as a New Frontier of Intelligence Workshop

See threads below for details!🧵👇

[Ying Wang @yingwww_]

Warm take: Your world model should never stop learning

Introducing AdaJEPA, an adaptive WM that plans, acts, and adapts in a closed loop. Every action leads to a new observation, and every transition refines the latent representation and prediction.

📝: https://t.co/85WWwn5QDS

[Jack Lu - COLM 26 @Jacklu_me]

Hot take: In-context learning is powerful, but one forward pass over the context is not enough.

Introducing Context Tuning: efficient LLM adaptation by refining memory representations rather than model weights, empowering ICL with inference-time optimization.

📍Tomorrow at @icmlconf! Jul 8 (Wed), 5 PM, Hall A #2510
📎https://t.co/r7wTBOlgei

[Prophet Arena @ProphetArena]

🔮 Forecast@#ICML2026 is almost here!

A full day of invited talks featuring @PTetlock, @SimonShaoleiDu, Nicole Kagan, Seth Blumberg, Scott Jeen, and Atlas Wang.

Plus industry sessions with @SusquehannaSIG, @Kalshi, @fleet_ai, and @_CausalLabs.

See you July 11 in Seoul! 🇰🇷

[Amelia Dai @ameliadai_]

Check more details in our paper! https://t.co/rsdFRvx3UV 🙏Many thanks to my collaborators @rteehas, @parsa_torabian, and espicially to my amazing advisor @mengyer @agentic_ai_lab!

[Amelia Dai @ameliadai_]

🧐Can we use the wisdom of crowds to calibrate LLM forecasters?

Introducing Beta-Bernoulli Calibrator (BBC): besides realized outcomes, we also learn from human forecast distributions, which reflect both crowd beliefs and consensus.

🔗https://t.co/dLz8ewMraw

[Amelia Dai @ameliadai_]

(5/5) How about a model’s confidence in its own forecast?

BBC captures epistemic uncertainty through the predicted Beta variance. Ideally, higher uncertainty should correspond to higher forecasting error, and BBC tracks Brier error much more reliably than verbalized confidence.