AdaFuse: Adaptive Ensemble Decoding with Test-Time Scaling for LLMs

Posted by klaurent Chicago, This Week: Astrophysics

Kavli Affiliate: Hsiaowen Chen| First 5 Authors: [#item_custom_name[1, [#item_custom_name[2, [#item_custom_name[3, [#item_custom_name[4, [#item_custom_name[5| Summary:Large language models (LLMs) exhibit complementary strengths arising from differences in pretraining data, model architectures, and decoding behaviors. Inference-time ensembling provides a practical way to combine these capabilities without retraining. However, existing ensemble approaches suffer from fundamental limitations. Most rely on fixed fusion […]


Continue.. AdaFuse: Adaptive Ensemble Decoding with Test-Time Scaling for LLMs

Chaining the Evidence: Robust Reinforcement Learning for Deep Search Agents with Citation-Aware Rubric Rewards

Posted by klaurent Peking, This Week: Astrophysics

Kavli Affiliate: Li Xin Li| First 5 Authors: [#item_custom_name[1, [#item_custom_name[2, [#item_custom_name[3, [#item_custom_name[4, [#item_custom_name[5| Summary:Reinforcement learning (RL) has emerged as a critical technique for enhancing LLM-based deep search agents. However, existing approaches primarily rely on binary outcome rewards, which fail to capture the comprehensiveness and factuality of agents’ reasoning process, and often lead to undesirable behaviors […]


Continue.. Chaining the Evidence: Robust Reinforcement Learning for Deep Search Agents with Citation-Aware Rubric Rewards

VENUS: Two Faint Little Red Dots Separated by $sim70,mathrmpc$ Hidden in a Single Lensed Galaxy at $zsim7$

Posted by klaurent Peking, This Week: Astrophysics

Kavli Affiliate: Kohei Inayoshi| First 5 Authors: Hiroto Yanagisawa, Hiroto Yanagisawa, , , | Summary:We report the identification of a pair of faint little red dots (LRDs), dubbed Red Eyes, in a strongly-lensed galaxy at $zsim7$ behind the PLCKG004.5-10.5 cluster, identified from the JWST Treasury program VENUS. Red Eyes are spatially resolved on the image […]


Continue.. VENUS: Two Faint Little Red Dots Separated by $sim70,mathrmpc$ Hidden in a Single Lensed Galaxy at $zsim7$

Slow mixing and emergent one-form symmetries in three-dimensional $mathbbZ_2$ gauge theory

Posted by klaurent Peking, This Week: Astrophysics

Kavli Affiliate: Li Xin Li| First 5 Authors: [#item_custom_name[1, [#item_custom_name[2, [#item_custom_name[3, [#item_custom_name[4, [#item_custom_name[5| Summary:Symmetry-breaking order at low temperatures is often accompanied by slow relaxation dynamics, due to diverging free-energy barriers arising from interfaces between different ordered states. Here, we extend this correspondence to classical topological order, where the ordered states are locally indistinguishable, so there […]


Continue.. Slow mixing and emergent one-form symmetries in three-dimensional $mathbbZ_2$ gauge theory

Discriminative-Generative Target Speaker Extraction with Decoder-Only Language Models

Posted by klaurent Peking, This Week: Astrophysics

Kavli Affiliate: Lile Wang| First 5 Authors: [#item_custom_name[1, [#item_custom_name[2, [#item_custom_name[3, [#item_custom_name[4, [#item_custom_name[5| Summary:Target speaker extraction (TSE) aims to recover the speech signal of a desired speaker from a mixed audio recording, given a short enrollment utterance. Most existing TSE approaches are based on discriminative modeling paradigms. Although effective at suppressing interfering speakers, these methods often […]


Continue.. Discriminative-Generative Target Speaker Extraction with Decoder-Only Language Models

Generalized Poincaré inequality for quantum Markov semigroups

Posted by klaurent Peking, This Week: Astrophysics

Kavli Affiliate: Lile Wang| First 5 Authors: [#item_custom_name[1, [#item_custom_name[2, [#item_custom_name[3, [#item_custom_name[4, [#item_custom_name[5| Summary:We prove a noncommutative $(p,p)$-Poincaré inequality for trace-symmetric quantum Markov semigroups on tracial von Neumann algebras, assuming only the existence of a spectral gap. Extending semi-commutative results of Huang and Tropp, our argument uses Markov dilations to obtain chain-rule estimates for Dirichlet forms […]


Continue.. Generalized Poincaré inequality for quantum Markov semigroups

The Molecular Structure of Thought: Mapping the Topology of Long Chain-of-Thought Reasoning

Posted by klaurent Chicago, This Week: Astrophysics

Kavli Affiliate: Hsiaowen Chen| First 5 Authors: [#item_custom_name[1, [#item_custom_name[2, [#item_custom_name[3, [#item_custom_name[4, [#item_custom_name[5| Summary:Large language models (LLMs) often fail to learn effective long chain-of-thought (Long CoT) reasoning from human or non-Long-CoT LLMs imitation. To understand this, we propose that effective and learnable Long CoT trajectories feature stable molecular-like structures in unified view, which are formed by […]


Continue.. The Molecular Structure of Thought: Mapping the Topology of Long Chain-of-Thought Reasoning

Two temperate Earth- and Neptune-sized planets orbiting fully convective M dwarfs

Posted by klaurent MIT, This Week: Astrophysics

Kavli Affiliate: Alan Levine| First 5 Authors: Madison G. Scott, Madison G. Scott, , , | Summary:As the diversity of exoplanets continues to grow, it is important to revisit assumptions about habitability and classical HZ definitions. In this work, we introduce an expanded ‘temperate’ zone, defined by instellation fluxes between $0.1<S/mathrmS_oplus<5$, thus encompassing a broader […]


Continue.. Two temperate Earth- and Neptune-sized planets orbiting fully convective M dwarfs

Special vs Essential

Posted by klaurent This Week: Astrophysics, Tokyo

Kavli Affiliate: Yukari Ito| First 5 Authors: Yukari Ito, Yukari Ito, , , | Summary:We show a correspondence between the compact exceptional curves and divisors on $G-rm Hilb(mathbfC^3)$ and some non-trivial irreducible representations of $G subset GL(n,C)$ which are special (or essential). Moreover, we provide an explicit construction of the small resolution of $G-rm Hilb(mathbfC^3)$ […]


Continue.. Special vs Essential

Robust Bilinear-Noise-Optimal Control for Gravitational-Wave Detectors: A Mixed LQG/$H_infty$ Approach

Posted by klaurent MIT, This Week: Astrophysics

Kavli Affiliate: Lee McCuller| First 5 Authors: Ian A. O. MacMillan, Ian A. O. MacMillan, , , | Summary:At its lowest frequencies, LIGO is limited by noise in its many degrees of freedom of suspended optics, which, in turn, introduce noise in the interferometer through their feedback control systems. Nonlinear interactions are a dominant source […]


Continue.. Robust Bilinear-Noise-Optimal Control for Gravitational-Wave Detectors: A Mixed LQG/$H_infty$ Approach