United States

Large-scale social simulators are essential for studying complex social patterns. Prior work explores hybrid methods to scale up simulations, combining large language models (LLM)-based agents with numerical agent-based models (ABM). However, this incurs high latency due to expensive memory retrieval and sequential ABM execution. To address this challenge, we propose GASim, a graph-accelerated hybrid multi-agent framework for large-scale social simulations. For core agents driven by LLM, GASim introduces Graph-Optimized Memory (GOM) to replace intensive LLM-based retrieval pipelines with lightweight propagation over a sparse memory graph. For the majority of ordinary agents, GASim employs Graph Message Passing (GMP), substituting sequential ABM execution with parallel updates by fine-grained feature aggregation and Graph Attention Network. We further introduce Entropy-Driven Grouping (EDG) that coordinates this hybrid partitioning, leveraging information entropy to dynamically identify emergent core agents situated in information-diverse neighborhoods. Extensive experiments show that GASim not only delivers a substantial 9.94× end-to-end speedup over the traditional hybrid framework but also consumes less than 20% of baseline tokens, significantly reducing costs while preserving strong alignment with real-world public opinion trends.

ACL 2026

 GASim: A Graph-Accelerated Hybrid Framework for Social Simulation

### Conference Welcome

Welcome to the third Cultural Evolution Society Conference, CES2021. In the conference, which will be held from June 9-11, 2021, about 180 researchers from East Asia, Oceania, Middle East, Europe, North and South America give presentations on their latest research advancing the evolutionary understanding of human and nonhuman animal culture.

### Schedule and Program

[![](https://assets.underline.io/uploads/markdown_image/1/image/1ef06ffd05f80957b4a5b685f63ee3ba.png)](https://www.chain.hokudai.ac.jp/CES2020/wp-content/uploads/2021/06/fullprogram.pdf)
[![](https://assets.underline.io/uploads/markdown_image/1/image/12285eb22580949635121d15d5b46325.png)](https://www.chain.hokudai.ac.jp/CES2020/wp-content/uploads/2021/06/abstracts.pdf)

**Program-at-glance**

![](https://assets.underline.io/uploads/markdown_image/1/image/c0d0d4c8fcdad49d12631574fd966ae8.png)

### About the conference 
 
**Help-desk:** 
Please first visit [Help-desk](https://underline.io/events/120/help-desk) where you can find the instructions and manuals for exploring the conference site. 

**Plenary sessions:**
Five leading researchers give a live talk. It is like a TV show. The audience can submit a question from the chat window. Chat messages are immediately delivered to speakers and moderators.

**Q&A sessions (parallel):**
Each presenter first gives a 2-minute live lightning talk followed by an 8-minute live discussion with the audience. It uses Zoom.

**Discussion sessions:** 
The audience can approach presenters and discuss their research. The sessions are held in a dedicated room in Gathertown.
 
**Gathertown:**
It is a platform dedicated to live virtual interactions.
Gather Town is available for all participants 24 hours during the conference.
You can meet and talk with anybody at a lounge or at a table in Gathertown.
Further Gather Town instruction and how to use short keys please find in our [Help desk page](https://underline.io/events/120/help-desk). 

**Videos and slides:**

* Plenary sessions and the Q&A sessions will be recorded and uploaded after the sessions.
* All the videos and slides are available for 60 days after the conference but only to registered ones. 

**Text-based chatting with the speakers:** 
You can chat with any speaker in the Q&A and discussion sessions. A chat window is available on each speaker’s page, and you can leave your questions there. 

**Announcement to the other participants:**
If you want to make an announcement to the other participants (e.g., “Hey, let’s meet at the lounge in Gathertown from 9pm CEST!), please use twitter (#CESConf2021).



### Code of conduct

See the full details of [the code of conduct](https://www.chain.hokudai.ac.jp/CES2020/code-of-conduct/).
&nbsp; 
 CES is dedicated to providing a welcoming and harassment-free conference experience for everyone, regardless of gender, gender identity and expression, sexual orientation, disability, physical appearance, body size, race, age, religion, nationality, or the presence of or care for dependents. 
We do not tolerate harassment of conference participants in any form. Conference participants violating these rules may be sanctioned or expelled from the conference at the discretion of the conference organizers.

&nbsp;
**Harassment can occur in-person or virtually, and includes, but is not limited to:**
* Comments (verbal, text, etc.) that reinforce social structures of domination related to gender, gender identity and expression, sexual orientation, disability, physical appearance, body size, race, age, religion, nationality
* Sexual images in public spaces (physical or virtual)
* Deliberate intimidation, stalking, or following
* Harassing photography or recording (including sharing of presentation materials without consent)
* Sustained disruption of talks or other events (including “Zoombombing”)
* Inappropriate physical contact or unwelcome sexual attention
* Advocating for, or encouraging, any of the above behaviour 

&nbsp;
 **Inclusiveness Committee**:
&nbsp;
 To facilitate implementing the code of conduct and foster a climate of inclusion, we have developed an inclusiveness committee. Below is their contact information:
 **Hiromi Matsumae**:	matsumae.hiromi.g@tokai.ac.jp
 **Patrick Savage**:	psavage@sfc.keio.ac.jp
 **Sarah Mathew**:	Sarah.Mathew@asu.edu
 **Gmail:** inclusivenessCES2021@gmail.com
 See [more details about reporting the harassment.](https://www.chain.hokudai.ac.jp/CES2020/code-of-conduct/)

### Organizing committee

**Honorary Conference Organizer:** [Kenichi Aoki (Meiji University)](https://sites.google.com/site/kenichi1aoki/home) 

**Chair:** [Masanori Takezawa (Hokkaido University)](https://lalalarakko.github.io)

**Scientific Committee:** [Kenichi Aoki (Meiji University)](https://sites.google.com/site/kenichi1aoki/home), [Yasuo Ihara (the University of Tokyo)](https://researchmap.jp/read0209180?lang=en), [Hiromi Matsumae (Tokai University)](https://sites.google.com/site/hiromimatsumae/eng), [Naoko Matsumoto (Okayama University)](https://researchmap.jp/nmatsumoto_okayama-u), [Masako Myowa (Kyoto University)](https://myowa.educ.kyoto-u.ac.jp/en), [Hisashi Nakao (Nanzan University)](https://www.hisashinakao.com/), [Patrick E. Savage (Keio University)](https://sites.google.com/view/comp-music-lab/home)

**Local Organizing Committee:** [Yutaka Horita (Teikyo University)](https://sites.google.com/site/yhoritapage/), [Masanori Takezawa (Hokkaido University)](https://lalalarakko.github.io/), [Ayaka Takimoto (Hokkaido University)](https://lynx.let.hokudai.ac.jp/~takimoto/), [Kohei Tamura (Tohoku University)](https://researchmap.jp/7000017854)

**Inclusiveness Committee:** [Sarah Mathew (Arizona State University)](https://www.sarahmathew.net/), [Hiromi Matsumae (Tokai University)](https://sites.google.com/site/hiromimatsumae/eng), [Patrick E. Savage (Keio University)](https://sites.google.com/view/comp-music-lab/home)


![](https://assets.underline.io/uploads/markdown_image/1/image/967110a00d200a6b20d6f4fa6556ba13.png)

For attending the conference, you first need to register and pay fees from [the link here](https://www.chain.hokudai.ac.jp/CES2020/registration/). After the payment of the registration fees, you will receive a ticket, which is necessary for accessing the conference site. After the payment of the registration fees, you will receive a ticket, which is necessary for accessing the conference site

Registration fee

Cultural Evolution Society 2021

poster

It is my great pleasure and honor to welcome you to the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), held in San Diego, California, USA, July 2–7, 2026. ACL 2026 continues our fields tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.

It is exciting to see the many developments in the field reflected at our conference, such as the advancement of large language models (born out of our community) into many application directions. This is not without challenges, partly due to unsolved problems, and partly due to misuse of the technology. All this is reflected well at ACL 2026. The conference also reflects that the field is not only about advancing technology, but also the science of fundamentally understanding language and how it is use by many different speaker communities. 
[Full Message from General Chair Philipp Koehn](https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing) 
<html><button style="display: inline-flex; align-items: center; justify-content: center; white-space: nowrap; border-radius: 9999px; font-weight: bold; background: #7c3aed; color: white; font-family: 'Space Grotesk', sans-serif; height: 40px; font-size: 16px; padding: 0 20px; border: none; cursor: pointer" onclick="window.open('https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing')">Conference Handbook</button></html>

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to ACL 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://2026.aclweb.org/registration/) first.

**Online Registration Form**: https://acl.swoogo.com/acl2026

Registration Required

The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026) will take place in San Diego, California, from July 2 - 7, 2026.

Process Reward Models (PRMs), which assign fine-grained scores to intermediate reasoning steps within a solution trajectory, have emerged as a promising approach to enhance the reasoning quality of Large Language Models (LLMs).
However, most existing PRMs rely on a unidirectional left-to-right (L2R) evaluation scheme, which restricts their utilization of global context.
In light of this challenge, we propose a novel bidirectional evaluation paradigm, named $\mathbf{Bi}$directional $\mathbf{P}$rocess $\mathbf{R}$eward $\mathbf{M}$odel ($\mathbf{BiPRM}$).
BiPRM incorporates a parallel right-to-left (R2L) evaluation stream, implemented via prompt reversal, alongside the conventional L2R flow.
Then a gating mechanism is introduced to adaptively fuse the reward scores from both streams to yield a holistic quality assessment.
Remarkably, compared to the original PRM, BiPRM introduces only a 0.3\% parameter increase for the gating module, and the parallel execution of two streams incurs merely 5\% inference time latency. 
Our extensive empirical evaluations spanning diverse benchmarks, LLM backbones, PRM objectives and sampling policies demonstrate that BiPRM consistently surpasses unidirectional baselines, achieving an average relative gain of 10.6\% over 54 solution-level configurations and 37.7\% in 12 step-level error detection scenarios.
Generally, our results highlight the effectiveness, robustness and general applicability of BiPRM, offering a promising new direction for process-based reward modeling.

The Bidirectional Process Reward Model

As large language models (LLMs) improve, many applications are moving from a single LLM call to multi-agent systems. These systems often rely on either hand-designed or automatically optimized workflows with multiple verification and testing steps. While those extra steps can improve accuracy, they also increase latency and token costs. In practice, many queries do not need such heavy processing and can be handled well by a single strong agent.

To address this inefficiency, we propose LLM-as-Scheduler (LAS), a system that dynamically chooses the right workflow for each query. LAS uses a two-stage cascade: first, a lightweight gate quickly evaluates each agent’s output; then, an LLM-based scheduler uses query features and gate signals to make more detailed routing decisions. Experiments show that LAS cuts token usage by 43% and reduces end-to-end latency by more than 36%, while causing at most a 1.4 percentage-point drop in accuracy compared with a strong fixed workflow.

LLM-as-Scheduler: Agentic Workflow Dynamic Scheduling

Large Language Models (LLMs) have demonstrated impressive capabilities across a wide range of NLP tasks, but they remain fundamentally stateless, constrained by limited context windows that hinder long-horizon reasoning. Recent efforts to address this limitation often augment LLMs with an external memory bank, yet most existing pipelines are static and heuristic-driven, lacking a learned mechanism for deciding what to store, update, or retrieve. We present Memory-R1, a reinforcement learning (RL) framework that equips LLMs with the ability to actively manage and utilize external memory through two specialized agents: a Memory Manager that learns structured operations, including ADD, UPDATE, DELETE, and NOOP; and an Answer Agent that pre-selects and reasons over relevant entries. Both agents are fine-tuned with outcome-driven RL (PPO and GRPO), enabling adaptive memory management with minimal supervision. With only 152 training QA pairs, Memory-R1 outperforms strong baselines and generalizes across diverse question types, three benchmarks (LoCoMo, MSC, LongMemEval), and multiple model scales (3B–14B).

Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning

Recent advances in large language models (LLMs) have highlighted the effectiveness of chain-of-thought reasoning in symbolic domains such as mathematics and programming. However, our study shows that directly transferring such text-based reasoning paradigms to protein function understanding is ineffective: reinforcement learning mainly amplifies superficial keyword patterns while failing to introduce new biological knowledge, resulting in limited generalization. We argue that protein function prediction is a knowledge-intensive scientific task that fundamentally relies on external biological priors and computational tools rather than purely internal reasoning. To address this gap, we propose Protein Function Understanding Agent (PFUA), a tool-augmented protein reasoning agent that unifies problem decomposition, tool invocation, and grounded answer generation. Instead of relying on long unconstrained reasoning traces, PFUA integrates domain-specific tools to produce verifiable intermediate evidence. Experiments on four benchmarks demonstrate that PFUA consistently outperforms text-only reasoning models with an average performance improvement of 103%. We believe PFUA has the potential to become a standard paradigm for agentic reasoning in knowledge-intensive life science domains.

Interleaved Tool-Call Reasoning for Protein Function Understanding

Reliable failure detection and causal reasoning are critical in robotic manipulation, as their absence risks robot damage and endangers human safety.
Although recent Vision–Language Models (VLMs) are employed to attempt failure detection and causality reasoning, they typically make retrospective assessment only after task completion, and their reasoning accuracy is often limited.
To address these issues, we introduce RoboFailRing, which enables timely failure detection during task execution and enhances the reasoning accuracy of VLMs.
It achieves rapid failure detection by retrieving a pre-constructed failure memory and returning a similarity-based decision.
In addition, by providing grounded failure report to VLMs, it improves the accuracy of their reasoning about the failure causes and repair strategies.
We evaluate RoboFailRing on two large-scale simulated datasets comprising over 6,000 failure trajectories and covering 81 distinct manipulation tasks.
The results show that the average success rate of out-of-distribution failure detection reaches 80%, while the mean detection time is cut to roughly 50% of the baseline.
Moreover, evaluations on real-world systems show an average 35% gain in VLM failure-reasoning accuracy.
We make our code publicly available at: https://github.com/DynamicPoet/RoboFailRing.

 RoboFailRing: Retrieval-Augmented and Language Grounding Failure Detection for VLM-enabled Robotic Manipulation

Recent studies have identified "retrieval heads" in Large Language Models (LLMs) responsible for extracting information from input contexts. However, prior works largely rely on static statistics aggregated across datasets, identifying heads that perform retrieval on average. This perspective overlooks the fine-grained temporal dynamics of autoregressive generation. In this paper, we investigate retrieval heads from a dynamic perspective. Through extensive analysis, we establish three core claims: (1) Dynamism: Retrieval heads vary dynamically across timesteps; (2) Irreplaceability: Dynamic retrieval heads are specific at each timestep and cannot be effectively replaced by static retrieval heads; and (3) Correlation: The model's hidden state encodes a predictive signal for future retrieval head patterns, indicating an internal planning mechanism. We validate these findings on the Needle-in-a-Haystack task and a multi-hop QA task, and quantify the differences on the utility of dynamic and static retrieval heads in a Dynamic Retrieval-Augmented Generation framework. Our study provides new insights into the internal mechanisms of LLMs.

 Retrieval Heads are Dynamic

Large transformer-based language models dominate modern NLP, yet our understanding of how they encode linguistic information relies primarily on studies of early models like BERT and GPT-2. We systematically probe 25 models from BERT Base to Qwen2.5-7B focusing on two linguistic properties: lexical identity and inflectional features across 6 diverse languages. We find a consistent pattern: inflectional features are linearly decodable throughout the model, while lexical identity is prominent early but increasingly weakens with depth. Further analysis of the representation geometry reveals that models with aggressive mid-layer dimensionality compression show reduced steering effectiveness in those layers, despite probe accuracy remaining high. Pretraining analysis shows that inflectional structure stabilizes early while lexical identity representations continue evolving. Taken together, our findings suggest that transformers maintain inflectional features across layers, while trading off lexical identity for compact, predictive representations.

 Model Internal Sleuthing: Finding Lexical Identity and Inflectional Features in Modern Language Models

Guideline-following is increasingly important in compliance, customer support, and other regulated workflows, where correctness is defined by explicit rule systems rather than heuristics. Learning to follow guidelines is challenging because guidelines are interdependent: rules can trigger, suppress, or conflict with one another, while locally plausible responses may violate global constraints. Most existing methods treat guidelines as static text and rely on implicit reasoning or deeper decoding, making rule interactions and satisfaction status hard to observe and control. A more feasible approach is to model guideline execution with an explicit state that tracks evolving rule evidence across steps. However, conventional world models are a poor fit: they typically assume privileged feedback or well-defined transition dynamics, assumptions that do not hold when reasoning occurs purely in language space under ambiguous, text-defined constraints. As a solution, we propose RGCWM, a Rule-Grounded Causal World Model that builds an explicit state space from the guideline text itself. RGCWM represents rule applicability and satisfaction as a continuously updated evidence state, externalizes inter-rule dependencies as a causal structure, and plans at inference time by counterfactually evaluating candidate responses under model-estimated state transitions. Experiments show that this shift from implicit text reasoning to state-based reasoning enables stable, controllable execution of complex interacting rules across diverse domains.

Guidelines as Environments: A World Model Approach to Rule Following

Language model (LM) probability is not a reliable quality estimator, as natural language is ambiguous. When multiple output options are valid, the model's probability distribution is spread across them, which can misleadingly indicate low output quality. This issue is caused by two reasons: (1) LMs' final output activation is softmax, which does not allow multiple correct options to receive high probabilities simultaneuously and (2) LMs' training data is single, one-hot encoded references, indicating that there is only one correct option at each output step. We propose training a module for Quality Estimation on top of pre-trained LMs to address these limitations. The module, called Sigmoid Head, is an extra unembedding head with sigmoid activation to tackle the first limitation. To tackle the second limitation, during the negative sampling process to train the Sigmoid Head, we use a heuristic to avoid selecting potentially alternative correct tokens. Our Sigmoid Head is computationally efficient during training and inference. The probability from Sigmoid Head is notably better quality signal compared to the original softmax head. As the Sigmoid Head does not rely on human-annotated quality data, it is more robust to out-of-domain settings compared to supervised QE.

 Sigmoid Head for Quality Estimation under Language Ambiguity

We introduce **Doublespeak**, a simple in-context representation hijacking attack against language models. The attack works by systematically replacing a harmful keyword (e.g., *bomb*) with a benign token (e.g., *carrot*) across multiple in-context examples, provided as a prefix to a harmful request. We demonstrate that this substitution leads to the internal representation of the benign token converging toward that of the harmful one, effectively embedding the harmful semantics under a euphemism. As a result, superficially innocuous prompts (e.g., *"How to build a carrot?"*) are internally interpreted as disallowed instructions (*"How to build a bomb?"*), thereby bypassing the model's safety alignment. We use interpretability tools to show this semantic shift occurs progressively across layers. Doublespeak is optimization-free, broadly transferable across model families, and achieves strong success rates on closed-source systems, reaching 74% on Llama-3.3-70B-Instruct with a single-sentence context override. Our findings highlight a new attack surface in LM latent space, indicating that current alignment strategies are insufficient and should instead operate at the representation level.

Downloads

Next from ACL 2026

The Bidirectional Process Reward Model

Similar lecture

Agent-based models in cultural evolution and in opinion dynamics: a path toward CE social science?

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES