arXiv subjects

Yike Wang

Publications and source records attributed to Yike Wang.

At least 19 recordsLinked to original sources

arXiv2026

Efficient Test-Time Adaptation through Human-AI Interaction

AI agents are trained on population-scale data to encode broad capabilities spanning those of many practitioners. Yet the artifacts they produce rarely meet the personal bar professionals need to stake their reputation on. On realistic, open-ended tasks where success criteria are heterogeneous and insufficiently documented, individual expertise lives precisely in the elevation and departure from the average. In practice, iterative human-agent interaction surfaces criteria that users cannot fully specify up front, yet apply repeatedly across tasks. We argue this cross-session interaction data is a rich, underused signal for closing the gap to individual expertise. In this work, we propose test-time adaptation through human-agent interaction (TAHI), which integrates these signals into agent context and weights, and crystallizes each user's training and evaluation criteria via an evolving rubric module. We adapt agents to 30 individuals in two high-utility domains, writing and visual creation, on a total of 600 tasks. Our agents improve solo task success by 4.5-20.9% within only tens of tasks. Meanwhile, our evolving rubric module serves as a scalable annotation tool, creating evaluation rubrics that catch 16.0-22.3% more failures than those from LMs or humans alone. While agents are adapted towards individuals, we show these personalized agents also produce improvements in success of up to 8.8% that generalize across users.

cs.AI↗

arXiv2026

Rethinking the Evaluation of Harness Evolution for Agents

We revisit the evaluation of automatic harness evolution for LLM agents. Existing harness evolution methods use unit test cases to search for harness configurations and then report final performance on the same public benchmark. This protocol raises two fundamental concerns. First, harness evolution is itself an iterative search procedure that repeatedly evaluates and revises candidate harnesses using task feedback. As in agentic test-time scaling, it should therefore be compared with simple task-level search baselines under matched feedback and inference budgets to determine whether its gains arise from improved harness design or from additional search alone. Second, because the search and the final evaluation share the same benchmark, the reported gains risk overfitting to that specific task set. To address these concerns, we conduct an extensive evaluation comparing harness evolution with simple test-time scaling and discovery baselines under comparable feedback and inference budgets, and also evaluate evolved harnesses on held-out tasks to assess whether the discovered improvements generalize. Experiments on Terminal-Bench 2.1 with GPT-5.4 and Claude Opus 4.6 show that automatic harness evolution does not consistently outperform simple test-time scaling methods and exhibits limited generalization. Our results raise important questions about the effectiveness of automatic harness evolution and highlight the need for fairer evaluation protocols and benchmarks for automatic harness design. Our code is available at https://github.com/rethinking-harness-evolution.

cs.AI↗

arXiv2026

Scaling Participation in Modular AI Systems

Humanity is a mosaic of multifaceted talents and needs, and any truly intelligent AI must reflect that richness. Yet the LLMs used by all are built by the few -- a centralized market of monolithic AI models structurally ill-suited to capture the diversity of human knowledge, reasoning, and values. Here we introduce scaling participation, a new paradigm in which modular AI systems are built from the bottom up through the contributions of diverse stakeholders. Participants contribute small models trained on their own interests and priorities; these models then collaborate in modular frameworks as compositional AI systems. Participatory AI systems outperform monolithic LLMs by up to 15.4% across 15 tasks, such as reasoning and factuality, surpassing models larger than all contributed components combined. Further experiments show that participatory AI systems benefit from contributor diversity, substantially improve on each contributor's original priorities, and exhibit emergent capabilities that allow them to solve over 15% of problems where all individual models fail. Scaling participation provides a technical foundation for transitioning from the monolithic status quo toward an open, bottom-up, and collaborative AI future.

cs.AI↗

arXiv2026

EvoLM: Self-Evolving Language Models through Co-Evolved Discriminative Rubrics

Language models encode substantial evaluative knowledge from pretraining, yet current post-training methods rely on external supervision (human annotations, proprietary models, or scalar reward models) to produce reward signals. Each imposes a ceiling. Human judgment cannot supervise capabilities beyond its own, proprietary APIs create dependencies, and verifiable rewards cover only domains with ground-truth answers. Self-improvement from a model's own evaluative capacity is a reward source that scales with the model itself, yet remains largely untapped by current methods. We introduce EVOLM, a post-training method that structures this capacity into explicit discriminative rubrics and uses them as training signal. EVOLM trains two capabilities within a single language model in alternation: (1) a rubric generator producing instance-specific evaluation criteria optimized for discriminative utility, which maximizes a small frozen judge's ability to distinguish preferred from dispreferred responses; and (2) a policy trained using those rubric-conditioned scores as reward. All preference signals are constructed from the policy's own outputs via temporal contrast with earlier checkpoints, requiring no human annotation or external supervision. EVOLM trains a Qwen3-8B model to generate rubrics that outperform GPT-4.1 on RewardBench-2 by 25.7%. The co-trained policy achieves 69.3% average on the OLMo3-Adapt suite, outperforming policies trained with GPT-4.1 prompted rubrics by 3.9% and with the state-of-the-art 8B reward model SkyWork-RM by 16%. Overall, EVOLM demonstrates that structuring a model's evaluative capacity into co-evolving discriminative rubrics enables self-improvement without external supervision.

cs.AI↗

arXiv2026

MoCo: A One-Stop Shop for Model Collaboration Research

Advancing beyond single monolithic language models (LMs), recent research increasingly recognizes the importance of model collaboration, where multiple LMs collaborate, compose, and complement each other. Existing research on this topic has mostly been disparate and disconnected, from different research communities, and lacks rigorous comparison. To consolidate existing research and establish model collaboration as a school of thought, we present MoCo: a one-stop Python library of executing, benchmarking, and comparing model collaboration algorithms at scale. MoCo features 26 model collaboration methods, spanning diverse levels of cross-model information exchange such as routing, text, logit, and model parameters. MoCo integrates 25 evaluation datasets spanning reasoning, QA, code, safety, and more, while users could flexibly bring their own data. Extensive experiments with MoCo demonstrate that most collaboration strategies outperform models without collaboration in 61.0% of (model, data) settings on average, with the most effective methods outperforming by up to 25.8%. We further analyze the scaling of model collaboration strategies, the training/inference efficiency of diverse methods, highlight that the collaborative system solves problems where single LMs struggle, and discuss future work in model collaboration, all made possible by MoCo. We envision MoCo as a valuable toolkit to facilitate and turbocharge the quest for an open, modular, decentralized, and collaborative AI future.

cs.CL↗

arXiv2026

Time-consistent portfolio selection with monotone mean-variance preferences

We investigate time-inconsistent portfolio problems under a broader class of monotone mean-variance (MMV) preferences. Since the optimal strategies for MMV and mean-variance (MV) preferences coincide, the MMV optimal strategies at different initial times are necessarily time-inconsistent. To address this time-inconsistency, we consider Nash equilibrium controls of both open-loop and closed-loop types, and characterize them within a random parameter setting. The two control problems reduce to solving a flow of forward-backward stochastic differential equations and a system of extended Hamilton-Jacobi-Bellman equations, respectively. In particular, we derive semi-closed-form solutions for both types of equilibria under a deterministic parameter setting, and both solutions share the same representation, which is independent of the wealth state and the random path. We show that the investment amount under the MMV equilibrium exceeds that under the MV equilibrium, and the gap narrows over time. Furthermore, under a constant parameter setting, we find that the derived closed-loop Nash equilibrium control is a strong equilibrium strategy only when the interest rate is sufficiently large, whereas the derived open-loop Nash equilibrium control is necessarily a strong equilibrium strategy.

math.OC↗

arXiv2026

Strictly monotone mean-variance preferences with applications to portfolio selection

The monotone mean-variance (MMV) preference proposed by Maccheroni, et al. (Math. Finance 19(3): 487-521, 2009) fails to differentiate strictly dominant payoffs, which may cause inconsistency in portfolio decision-making. This paper introduces a broader class of strictly monotone mean-variance (SMMV) preferences and demonstrates its applications to portfolio selection problems. For the single-period portfolio problem under the SMMV preference, we derive the gradient condition for the optimal strategy, and investigate its association with the optimal mean-variance (MV) static strategy. We reduce the problem to solving a set of linear equations by analyzing the saddle point of some minimax problem. And results show that the optimal SMMV, MMV and MV strategies differ significantly in the single-period problem. Furthermore, we conduct numerical experiments and compare our results with those of Maccheroni, et al. (Math. Finance 19(3): 487-521, 2009). The findings indicate that our SMMV preferences provide a more rational basis for assessing given prospects. For the continuous-time portfolio problem under the SMMV preference, we consider continuous price processes with random coefficients, and establish a novel approach based on a general convex duality analysis to derive the optimal strategy. Interestingly, we find that the optimal strategies for SMMV, MMV and MV preferences coincide under a certain condition, and provide a classical microeconomic interpretation for this condition. We also characterize the optimal SMMV portfolio strategies relying on stochastic control techniques to facilitate potential extensions and refinements in future research.

q-fin.MF↗

arXiv2026

Equilibrium strategies for stochastic control problems with higher-order moments and applications to portfolio selection

In this paper we derive a novel characterization result for time-consistent stochastic control problems with higher-order moments, originally formulated by Wang et al. [SIAM J. Control. Optim., 63 (2025), 1560--1589], and newly explore many solvable instances including a mean-variance-excess kurtosis portfolio selection problem. By improving an asymptotic result of the variational process for the uniform boundedness and integrability properties, we obtain both the sufficiency and necessity of an equilibrium condition for an open-loop Nash equilibrium control (ONEC). This condition is simply formulated by the diagonal processes of a flow of backward stochastic differential equations (BSDEs) whose data do not necessarily satisfy the usual square-integrability condition. In particular, for linear controlled dynamics with deterministic parameters, we show that the ONEC can be derived by solving a polynomial algebraic equation under a class of nonlinear objective functions. Interestingly, the mean-variance equilibrium strategy is an ONEC for our general higher-order moment problem if and only if a homogeneity condition holds. Additionally, in the case with random parameters, we characterize the ONEC by finitely many BSDEs with a recurrence relation. As an intuitive illustration, the solution to the mean-variance-skewness problems is given by a quadratic BSDE.

math.OC↗

arXiv2026

Small Reward Models via Backward Inference

Reward models (RMs) play a central role throughout the language model (LM) pipeline, particularly in non-verifiable domains. However, the dominant LLM-as-a-Judge paradigm relies on the strong reasoning capabilities of large models, while alternative approaches require reference responses or explicit rubrics, limiting flexibility and broader accessibility. In this work, we propose FLIP (FLipped Inference for Prompt reconstruction), a reference-free and rubric-free reward modeling approach that reformulates reward modeling through backward inference: inferring the instruction that would most plausibly produce a given response. The similarity between the inferred and the original instructions is then used as the reward signal. Evaluations across four domains using 13 small language models show that FLIP outperforms LLM-as-a-Judge baselines by an average of 79.6%. Moreover, FLIP substantially improves downstream performance in extrinsic evaluations under test-time scaling via parallel sampling and GRPO training. We further find that FLIP is particularly effective for longer outputs and robust to common forms of reward hacking. By explicitly exploiting the validation-generation gap, FLIP enables reliable reward modeling in downscaled regimes where judgment methods fail. Code available at https://github.com/yikee/FLIP.

cs.CL↗

arXiv2026

MentorCollab: Large-to-Small Inference-Time Mentorship for Concise Reasoning in Language Models

Large reasoning models (LRMs) have demonstrated impressive reasoning capabilities, but their solutions are often verbose and computationally expensive, and taxing for users to read. In contrast, small language models (SLMs) produce concise outputs with lower inference costs, yet they frequently struggle on challenging multi-step reasoning tasks. Existing inference-time collaboration methods attempt to bridge this gap through imitation, encouraging SLMs to follow the reasoning process of LRMs. However, the student often inherits the mentor's overthinking, producing long and reflective reasoning chains while still falling short in accuracy. We propose MentorCollab, a collaboration method based on mentorship: the SLM remains the primary generator and consults the LRM only when additional reasoning support is needed. At sparsely sampled token positions, we probe for divergence between the two models and use a lightweight verifier to decide whether the SLM should follow a short lookahead segment from its mentor or continue on its own. Across 15 SLM-LRM pairs and 3 domains, our method achieves an average gain of 3.0%, with improvements of up to 8.0% in 12 settings. The resulting traces remain shorter than the mentor's, using only a small fraction of its tokens. These results demonstrate that selective, verified mentorship can boost reasoning accuracy while preserving concise generation.

cs.CL↗

arXiv2025

CGM-Led Multimodal Tracking with Chatbot Support: An Autoethnography in Sub-Health

Metabolic disorders present a pressing global health challenge, with China carrying the world's largest burden. While continuous glucose monitoring (CGM) has transformed diabetes care, its potential for supporting sub-health populations -- such as individuals who are overweight, prediabetic, or anxious -- remains underexplored. At the same time, large language models (LLMs) are increasingly used in health coaching, yet CGM is rarely incorporated as a first-class signal. To address this gap, we conducted a six-week autoethnography, combining CGM with multimodal indicators captured via common digital devices and a chatbot that offered personalized reflections and explanations of glucose fluctuations. Our findings show how CGM-led, data-first multimodal tracking, coupled with conversational support, shaped everyday practices of diet, activity, stress, and wellbeing. This work contributes to HCI by extending CGM research beyond clinical diabetes and demonstrating how LLM-driven agents can support preventive health and reflection in at-risk populations.

cs.HC↗

arXiv2025

Heterogeneous Swarms: Jointly Optimizing Model Roles and Weights for Multi-LLM Systems

We propose Heterogeneous Swarms, an algorithm to design multi-LLM systems by jointly optimizing model roles and weights. We represent multi-LLM systems as directed acyclic graphs (DAGs) of LLMs with topological message passing for collaborative generation. Given a pool of LLM experts and a utility function, Heterogeneous Swarms employs two iterative steps: role-step and weight-step. For role-step, we interpret model roles as learning a DAG that specifies the flow of inputs and outputs between LLMs. Starting from a swarm of random continuous adjacency matrices, we decode them into discrete DAGs, call the LLMs in topological order, evaluate on the utility function (e.g. accuracy on a task), and optimize the adjacency matrices with particle swarm optimization based on the utility score. For weight-step, we assess the contribution of individual LLMs in the multi-LLM systems and optimize model weights with swarm intelligence. We propose JFK-score to quantify the individual contribution of each LLM in the best-found DAG of the role-step, then optimize model weights with particle swarm optimization based on the JFK-score. Experiments demonstrate that Heterogeneous Swarms outperforms 15 role- and/or weight-based baselines by 18.5% on average across 12 tasks. Further analysis reveals that Heterogeneous Swarms discovers multi-LLM systems with heterogeneous model roles and substantial collaborative gains, and benefits from the diversity of language models.

cs.CL↗

arXiv2025

Don't Throw Away Your Pretrained Model

Alignment training has tradeoffs: it helps language models (LMs) gain in reasoning and instruction following but might lose out on skills such as creativity and calibration, where unaligned base models are better at. We aim to make the best of both worlds through model collaboration, where different models in the training pipeline collaborate and complement each other. Since LM responses feature interleaving skills that favor different models, we propose Switch Generation, where pretrained and aligned model versions take turns to ``speak'' in a response sequence. Specifically, we train a switcher LM by learning from outcomes of choosing different models to generate the next segment across diverse queries and contexts. At inference time, the switcher LM guides different model checkpoints to dynamically generate the next segment where their strengths are most needed. Extensive experiments with 8 model collaboration baselines and 18 datasets show that 1) model collaboration consistently outperforms individual models on 16 out of 18 tasks, and 2) Switch Generation further outperforms baselines by 12.9% on average. Further analysis reveals that Switch Generation discovers compositional skills to solve problems where individual models struggle and generalizes to unseen models and tasks, reusing and repurposing by-products in expensive model training pipelines that are otherwise discarded.

cs.CL↗

arXiv2025

Closed-loop strong equilibria for time-inconsistent control problems with higher-order moments

In this paper, we study closed-loop strong equilibrium strategies for the time-inconsistent control problem with higher-order moments formulated by [Wang et al. SIAM J. Control. Optim., 63 (2025), 1560--1589]. Since time-inconsistency makes the dynamic programming principle inapplicable, the problem is treated as a game between the decision maker and her future selves. As a time-consistent solution, the previously proposed Nash equilibrium control is merely a stationary point and does not necessarily reach a maximum in the game-theoretical prospective. To address this issue, we consider the strong equilibrium strategy, from which any spike deviation will be worse off. We derive sufficient conditions for strong equilibrium strategies by expanding the objective function corresponding to the spike deviation with respect to the variational factor up to higher orders, and simplify the result by exploiting the partial differential equations that a Nash equilibrium control satisfies. Next, we examine whether the derived Nash equilibrium control is a strong equilibrium strategy. In particular, we provide a sufficient condition for the case with linear controlled equations. When the model parameters are constant, we find that the derived Nash equilibrium control for the mean-variance problem must be a strong equilibrium strategy, while that for the mean-variance-skewness-kurtosis problem is not necessarily a strong equilibrium strategy.

math.OC↗

arXiv2025

QoS-Aware Integrated Sensing, Communication, and Control with Movable Antenna

Integrated sensing, communication, and control (ISCC) has emerged as a key enabler for low-altitude wireless networks with enhanced adaptability through resource allocation co-design and intelligent environment awareness. However, dynamic interference and channel attenuation constrain the potential of the ISCC system. To address this challenge, we propose a novel movable antenna-empowered ISCC system. An achievable data rate maximization problem is formulated while guaranteeing the sensing and control quality-of-service (QoS) by optimizing the positions of the antennas and the beamforming strategy for communication, sensing, and control co-design. An efficient alternating optimization (AO)-based algorithm is proposed to solve the highly coupled non-convex problem. Numerical results demonstrate that the proposed AO-based algorithm achieves substantial gains in the achievable data rate and the control QoS compared with benchmark schemes.

cs.IT↗

arXiv2025

Data Swarms: Optimizable Generation of Synthetic Evaluation Data

We propose Data Swarms, an algorithm to optimize the generation of synthetic evaluation data and advance quantitative desiderata of LLM evaluation. We first train a swarm of initial data generators using existing data, and define various evaluation objectives to reflect the desired properties of evaluation (e.g., generate more difficult problems for the evaluated models) and quantitatively evaluate data generators. We then employ particle swarm optimization to optimize the swarm of data generators, where they collaboratively search through the model parameter space to find new generators that advance these objectives. We further extend it to Adversarial Swarms, where the data generator swarm generates harder data while the test taker model swarm learns from such data, co-evolving dynamically for better data and models simultaneously. Extensive experiments demonstrate that Data Swarms outperforms eight data generation baselines across five evaluation objectives, while Adversarial Swarms produce more robust learning of synthetic data and stronger generalization. Further analysis reveals that Data Swarms successfully optimizes compositions of multiple evaluation objectives and generalizes to new off-the-shelf LLMs, unseen at optimization time.

cs.CL↗

arXiv2025

Model Swarms: Collaborative Search to Adapt LLM Experts via Swarm Intelligence

We propose Model Swarms, a collaborative search algorithm to adapt LLMs via swarm intelligence, the collective behavior guiding individual systems. Specifically, Model Swarms starts with a pool of LLM experts and a utility function. Guided by the best-found checkpoints across models, diverse LLM experts collaboratively move in the weight space and optimize a utility function representing model adaptation objectives. Compared to existing model composition approaches, Model Swarms offers tuning-free model adaptation, works in low-data regimes with as few as 200 examples, and does not require assumptions about specific experts in the swarm or how they should be composed. Extensive experiments demonstrate that Model Swarms could flexibly adapt LLM experts to a single task, multi-task domains, reward models, as well as diverse human interests, improving over 12 model composition baselines by up to 21.0% across tasks and contexts. Further analysis reveals that LLM experts discover previously unseen capabilities in initial checkpoints and that Model Swarms enable the weak-to-strong transition of experts through the collaborative search process.

cs.CL↗

arXiv2025

ScienceMeter: Tracking Scientific Knowledge Updates in Language Models

Large Language Models (LLMs) are increasingly used to support scientific research, but their knowledge of scientific advancements can quickly become outdated. We introduce ScienceMeter, a new framework for evaluating scientific knowledge update methods over scientific knowledge spanning the past, present, and future. ScienceMeter defines three metrics: knowledge preservation, the extent to which models' understanding of previously learned papers is preserved; knowledge acquisition, how well scientific claims from newly introduced papers are acquired; and knowledge projection, the ability of the updated model to anticipate or generalize to related scientific claims that may emerge in the future. Using ScienceMeter, we evaluate the scientific knowledge of LLMs through claim judgment and generation tasks on a curated dataset across ten domains. We evaluate five representative knowledge update approaches and find that the best-performing knowledge update methods can preserve only 85.9% of existing knowledge, acquire 71.7% of new knowledge, and project 37.7% of future knowledge, underscoring that developing robust scientific knowledge update mechanisms is both crucial and challenging.

cs.CL↗