Summarization
-
Can One Adapted Model Do It All? Fine-Tuning Strategy Selection for Customer Support LLMs
2026-09-23 02:46:50Md Tahmid Rahman Laskar, Xue-Yong Fu, Shashi Bhushan TNAbstract
Production customer-support systems often require LLMs to support multiple skills, such as intent classification, question answering, summarization, or tool-use decisions. A central deployment question is whether these skills should be handled by separate task-specialist models or by a single model trained through multi-task training, sequential updates, or model merging. We study this question using thirteen models spanning five families (Qwen3, Qwen3.5, Gemma-3, Llama-3.1, and Mistral) from 0.6B to 32B parameters across eight customer-support datasets, spanning four public and four proprietary datasets with approximately 74.5k training and 8.7k evaluation samples. Under a fixed training protocol, we train more than 200 checkpoints. Our experiments reveal that multi-task full fine-tuning is the strongest operational default at every model size we test. Specialist models are strong on their target tasks but often degrade sharply off-task, making reliable routing important. Sequential Low-Rank Adaptation (LoRA) preserves earlier skills better than sequential full fine-tuning, while merging a specialist with its base model improves off-task robustness with limited same-task loss for larger models. We conclude with practical guidelines for selecting fine-tuning strategies in real-world settings.
Abstract (translated)
URL
https://arxiv.org/abs/2609.27262
PDF
-
Distilling Sequential Computation in Transformer Language Models
2026-09-23 02:02:57Zixuan Lan, Jessica Yang, Yanhong Li, Karen Livescu, Jiawei ZhouAbstract
Transformer language models process sequences token by token in an autoregressive manner, making growing contexts increasingly expensive. Yet many adjacent token spans are highly predictable or frequently occur as stable units, suggesting that their representations may be compressible. We introduce a method for distilling sequential computation by replacing spans of input tokens with collapsed representations, computed on the fly by a lightweight merge module. This module generates a single surrogate embedding from a sequence of static token embeddings that captures the functional role of the multiple tokens, allowing pretrained models to operate on compressed inputs without architectural changes or re-training. We apply this approach during inference to compress both prompts and intermediate decoding steps, using a rollback mechanism to substitute stored multi-token KV cache entries with their single-step surrogates. Experiments across diverse models show that the merge module can be used to reduce effective sequence length by up to 40% with minimal accuracy degradation across language modeling evaluations and downstream tasks, including question answering, summarization, commonsense reasoning, and long-form mathematical reasoning. Additional lightweight adaptation of the merge module further improves the accuracy-compression trade-off in selected settings. These results demonstrate that sequential token computation in Transformers can be effectively approximated through condensed surrogate representations that approximate the original behavior without model updating.
Abstract (translated)
URL
https://arxiv.org/abs/2609.27233
PDF
-
LexLattice: Multilingual Extractive Summarization via Neural Cellular Automata on Document Hierarchies
2026-09-22 20:25:30Sujay Uday Rittikar, Sheela RamannaAbstract
Faithfulness is a central concern in legal text summarization, which motivates extractive approaches that select verbatim content traceable to its source. Such methods typically rank paragraphs or other structural units in isolation, yet give little attention to consolidating evidence that is distributed across, and shares salience between, distant parts of a document. We introduce LexLattice, an extractive summarizer that reifies a legal act's hierarchy as a two-dimensional semantic lattice and consolidates over it with a masked 2D neural cellular automata before selection. LexLattice attains state-of-the-art ROUGE across all 24 languages of EUR-Lex-Sum in both multilingual and cross-lingual settings, surpassing instruction-tuned baselines with billions of parameters, despite concentrating all trainable capacity in a 1.8M parameter consolidator over a frozen multilingual encoder. A consolidator trained only on high-resource languages further transfers to unseen languages with near-lossless retention (0.99), indicating that the model operates on language-agnostic semantic geometry rather than surface form. Our results position explicit consolidation over document structure as a compact and traceable alternative to scale for multilingual legal summarization.
Abstract (translated)
URL
https://arxiv.org/abs/2609.27032
PDF
-
Identity-Centric Video Summarization via Hierarchical Fusion of Biometric, Appearance, and 3D Body Features
2026-09-22 08:05:09Milad Mirjalili, Enrique Alegre Guti\'errez, Eduardo Fidalgo Fern\'andez, V\'ictor Gonz\'alez Castro, Roc\'io Alaiz Rodr\'iguez, Manuel Castej\'on LimasAbstract
This work presents a video summarization algorithm based on multi-object tracking and person reidentification. We integrate facial embeddings, 3D body-shape features, and visual appearance into a unified tracking framework. These representations enable hierarchical identity assignment and tracking through bidirectional anchoring, which robustly recovers trajectories under severe occlusion or low visual quality. From these stable trajectories, we generate a compact set of summaries for each identity. We select keyframes using a multi-factor weighting scheme that optimizes biometric clarity, social interaction, and motion dynamics, while Adaptive Non-Maximum Suppression ensures temporal diversity. Evaluation on a custom dataset demonstrates tracking stability, achieving an IDF1 of 97.89% and a MOTA of 95.79%. Compared to Top-K selection, our algorithm also increases visual diversity by 146%, temporal coverage by 89%, and information retrievability by 3.5%.
Abstract (translated)
URL
https://arxiv.org/abs/2609.25837
PDF
-
Speech-to-SOAP: End-to-End Summarization of Medical Dialogues: KIT@BeTraC 2026
2026-09-21 10:04:09Enes Yavuz Ugan, Fabian Retkowski, Yuka Ko, Thai-Binh Nguyen, Maike Z\"ufle, Jan Niehues, Alexander WaibelAbstract
With the advent of Large Language Models and its instruction following capabilities a promising application is the task of summarization. Within this domain of task the extractive sub-task of clinical protocolling has emerged as a topic of particular interest as it can significantly reduce the downtime and protocolling burden of health-care workers thus enabling them to focus on their core work helping humans. A further step towards automation is the direct generation of clinical notes from speech without intermediate transcripts, reducing processing time while preserving information such as coughing or other paralinguistic cues that may be lost in transcript-based systems. To this end, we present KIT's submission to this years BeTraC challenge in the lightweight track. Our main contribution is a scalable data augmentation pipeline that unifies heterogeneous medical dialogue datasets through synthetic speech generation and automatically generated SOAP supervision, enabling robust adaptation of a speech foundation model for end-to-end speech-to-SOAP generation.
Abstract (translated)
URL
https://arxiv.org/abs/2608.24327
PDF
-
When Evidence Conflicts: Reliability-aware Meta-review Generation
2026-09-21 02:50:49Xinzhe Wang, Fei Tao, Jiang Xie, Hong Yu, Ye WangAbstract
Generating coherent meta-reviews from multiple peer reviews is challenging when reviewer evidence conflicts and varies in reliability. Existing approaches typically formulate meta-review generation as a multi-document summarization task and aggregate reviewer feedback uniformly, making it difficult to determine which opinions should be prioritized under disagreement. In this paper, we study meta-review generation through reliability-aware evidence aggregation. Our framework first extracts aspect-level opinions from peer reviews and identifies conflicting evidence within each aspect. It then estimates opinion-level support and review-level quality to measure evidence reliability. Based on these signals, the framework assigns reliability-aware weights to reviewer feedback, enabling the generator to prioritize better-supported arguments while preserving diverse perspectives. Experiments demonstrate that our method consistently improves meta-review generation over strong baselines on both automatic and human evaluations, with clear gains in conflict recognition and resolution under high-conflict review scenarios. The code and implementation details are publicly available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2609.24028
PDF
-
Preserving What Matters: Semantic Scaffolds Beyond Saturation in Summarization Evaluation
2026-09-18 21:43:14Nikhil Reddy Pottanigari, Ramin Fahimi, Noah Bolger, Sepideh Kharaghani, Ying ZhangAbstract
Summarization ships in countless production systems, making model selection a routine decision that depends on measuring summary quality. Existing metrics struggle to support this: ROUGE captures only surface overlap, while LLM-as-judge scores saturate to near-identical values that fail to rank models effectively. We observe this saturation across three public datasets, two proprietary datasets, and multilingual settings. Motivated by this, we introduce Semantic Scaffold, an evaluation framework that extracts a hierarchical representation of facts, questions, and entity attributes from a source text, labeling each as a main point or supporting detail, and reusing this structure as a fixed reference for scoring summaries. From this representation, we derive three diagnostic metrics: Fact Preservation Score (FPS), Question Preservation Score (QPS), and Entity Preservation Score (EPS), designed to reward the preservation of essential information while penalizing detail overload, and position them as interpretable diagnostics that remain informative where holistic axes collapse. Finally, we analyze four recurring failure modes of ROUGE and LLM-as-judge scores, demonstrating that scaffold-based evaluation remains informative where conventional metrics collapse.
Abstract (translated)
URL
https://arxiv.org/abs/2609.22603
PDF
-
Toward Personalized Sleep Guidance from Wearable Data Using Language Models
2026-09-18 18:22:20Yusheng Tan, Running Zhao, Sofia Angel, Ninghui Hao, Ash Arian, Nikita N. Dulin, Jay Lin, Ou Zhu, Faiza Shaik, Xinxing Yang, Bonnie W. Leung, Katie Roster, Arlene Ruiz de Luzuriaga, Kenneth Lee, Alejandra Lastra, Habibul Ahsan, Guihong WanAbstract
Sleep monitoring using wearable data has shown promise for personal health, yet large language model (LLM)-based summarization and question answering remain insufficient for personalized sleep guidance. Training specialized models, however, often requires costly expert annotation. Moreover, privacy and accessibility concerns motivate lightweight, local deployment for end users. We present a two-stage framework to address these challenges. Specifically, in Stage~1, a multi-agent LLM pipeline reasons structured sleep guidance from unannotated wearable records, enabling scalable dataset construction. Stage~2 distills guidance reasoning trajectories into small language models (SLMs) through supervised fine-tuning and integrates a training-free Best-of-$N$ selection strategy to enhance inference. Experimental results demonstrate our method outperforms commercial general and medical LLMs and open-source models. Human evaluation further supports the quality of the generated guidance and the feasibility of personalized sleep guidance with SLMs.
Abstract (translated)
URL
https://arxiv.org/abs/2609.22463
PDF
-
When Does Reasoning Help in Machine Translation? A Hierarchical Analysis of LRM Reasoning Traces
2026-09-18 02:43:26Yuxiang Liu, Jiaming Luo, Eleftheria Briakou, Colin CherryAbstract
Large Reasoning Models increasingly use intermediate traces for machine translation, but it remains unclear when such reasoning helps or hurts. We analyze reasoning traces across models, languages, domains, and datasets, focusing on reasoning language, length, and structure. We find that the best reasoning language is model-specific, reasoning length has a non-monotonic relationship with quality, and traces exhibit recurring functional patterns. To uncover these patterns, we introduce Hierarchical Meta-Summarization (HMS), a scalable framework that induces coarse- and fine-grained reasoning structures without predefined taxonomies. HMS reveals a shared organization--understanding/planning, translating/drafting, and refining/verifying--alongside domain-specific variation. Our results suggest that MT reasoning should be controlled in a model-aware, length-aware, and pattern-aware manner rather than uniformly encouraged.
Abstract (translated)
URL
https://arxiv.org/abs/2609.21247
PDF
-
An Empirical Study of Harness Design for Coding Agents
2026-09-17 17:58:07Run-Ze Fan, Zihao Zhang, Simin Ma, Yebowen Hu, Shouju Wang, Kaiqiang Song, Fei Liu, Hamed Zamani, Xiaoyang WangAbstract
Coding harnesses shape how autonomous coding agents translate model capabilities into long-horizon software-engineering performance, yet existing work typically evaluates harnesses as monolithic systems, leaving the effectiveness of individual components unclear. To enable component-level comparisons, we study this question with a lightweight coding harness whose execution loop is fixed while three components are varied: planning, action space, and context management. Across four models evaluated on SWE-Bench Verified and Terminal-Bench 2.1, we evaluate 176 matched settings spanning five context-management strategies, four context-window budgets, and targeted ablations of planning and action space. We find that: (1) Context management becomes increasingly valuable as the context-window budget tightens, with most of its benefit coming from preventing context-overflow failures. (2) Staging rule-based elision before LLM-based summarization provides the strongest overall efficiency among the context-management strategies, whereas making elided content recoverable adds machinery that models rarely use and yields no accuracy gain. (3) Planning shifts from an accuracy scaffold for weaker models to a cost saver for stronger models, with little change in accuracy. (4) Predefined tools improve performance for models with weaker bash proficiency, whereas bash-capable models can operate effectively with a bash-only interface and achieve substantially lower cost, especially on command-line-centric tasks. Trajectory-level analysis explains these effects: context management extends execution trajectories without substantially altering agent behavior, planning changes where trajectories stop, and the action space changes the granularity at which code is written. These findings inform model- and budget-aware harness design and provide a modular framework for evaluating future harness components.
Abstract (translated)
URL
https://arxiv.org/abs/2609.20804
PDF
-
Summarization Bias: The Directional Collapse of Objective Projection into Told-Mode Labels in Large Language Models --- A Conceptual Framework and Registered Test Protocol
2026-09-17 17:09:58Levent BulutAbstract
This paper introduces and operationalizes summarization bias: a proposed systematic tendency of large language models (LLMs) to represent narrative meaning as an abstract summary label rather than as the reconstructable inferential structure that produces it. Within the Bulut Doctrine, narrative effect is theorized along a told-shown axis: in told mode, emotional and informational content is declared explicitly and requires little reader reconstruction; in shown mode, that content is suppressed at the surface and must be reconstructed from physical cues and indirection (Objective Projection). Shown mode is the higher-load condition the doctrine is designed to measure. The claim is that LLMs fail along this axis in a specific direction. Summarization bias is hypothesized to operate in two regimes: (i) a generative regime, in which a model asked to render an emotion through Objective Projection defaults to declaring it instead; and (ii) an evaluative regime, in which a model judging narrative quality rewards told-mode explicitness and under-detects shown-mode suppression. The evaluative regime is the more consequential, since LLMs increasingly serve as judges and reward models, and a directional bias toward told mode would impose a selection pressure degrading prose toward flat declaration. This report does not claim the bias is validated. It defines the construct, situates it against LLM-as-judge biases, rereads a completed independent reliability study as directional evidence consistent with it, and pre-registers a two-regime test with decision rules under which the construct would be abandoned.
Abstract (translated)
URL
https://arxiv.org/abs/2609.20712
PDF
-
CliniCIRCA: A Modular LLM Framework for Constructing Longitudinal Mental Health Patient Journeys from Raw EHR Narratives
2026-09-17 02:13:21Aiwei Ivy Zhang, Nimra Ishfaq, Mohit Chandra, Santiago Alvarez Lesmes, Adam Coscia, Khatiya Chelidze Moon, Xiaohan Ding, Munmun De ChoudhuryAbstract
In mental health care, reasoning over patient journeys is a key task for clinicians. Yet these journeys, encompassing a longitudinal progression of biological, psychological, and social events, are often spread across disparate unstructured text narratives, making temporal recovery challenging. We present CliniCIRCA, a multi-stage LLM framework for Calendar-anchored, Imprecision-aware Reconstruction of Clinical Annals. To our knowledge, CliniCIRCA is the first to temporally classify clinical events across unstructured discharge summaries without event-level timestamps. From 14,882 MIMIC-III mental health admissions, we first construct a benchmark of 52 discharge summaries on which CliniCIRCA produces 15,891 temporally tagged events. After correcting 629 errors based on a clinician-in-the-loop evaluation, we produce verified gold-standard labels. Finally, the corrected timelines drive a temporally grounded summarization stage that compresses each source 1.52 times into a date-grouped chronological record. We then scale the framework to generate 1,000 silver-standard timelines and evaluate them as training data. Compared with zero- and few-shot prompting, instruction tuning generally improves five open-weight models on event extraction, temporal tagging, and summarization across silver and clinician-verified evaluations.
Abstract (translated)
URL
https://arxiv.org/abs/2609.19585
PDF
-
SPEAR NeXT Causal Latent Forecasting Across Multiple Horizons for Spectral Temporal Earth Representation Learning
2026-09-15 09:04:10Rajiv Ranjan, Udaiveer Singh, Shashank Tamaskar, Dharmendra SaraswatAbstract
Earth observation is inherently dynamic, yet temporal information in many foundation models is learned through reconstruction, invariance, or retrospective sequence summarization. SPEAR NeXT is introduced as a compact pixel-wise multimodal spectral temporal foundation model in which temporal self supervision is formulated as past only, multi horizon latent Earth state prediction. Instantaneous states are first encoded by the pretrained SPEAR model from optical, radar, and environmental observations into compact 32 dimensional embeddings. Their temporal evolution is then modeled by a causally masked Trans former that predicts multiple future latent states from pre ceding observations. Relative temporal order is represented using Rotary Position Embeddings, while month and year embeddings encode seasonal phase and interannual con text.
Abstract (translated)
URL
https://arxiv.org/abs/2609.16871
PDF
-
TIAO: Token Importance-Aware Policy Optimization for Text Summarization
2026-09-15 07:23:07Qixiu Li, Chenlong Bao, Xiang Zhu, Xiaoyong Li, Ruixin Cao, Shukai Chen, Zhenxiong ZhouAbstract
Text summarization requires models to condense content while preserving key qualities such as consistency and coherence. Large language models (LLMs) have shown strong performance on this task and can be further improved through reinforcement learning (RL). However, most existing methods apply reward signals directly to undifferentiated token sequences, overlooking the varying importance of individual tokens to word and sentence level quality in summarization. In this paper, we propose Token Importance-Aware Policy Optimization (TIAO), a novel reinforcement learning strategy that explicitly leverages token-importance awareness. Specifically, TIAO identifies core tokens based on token dependency and reweights a trajectory's advantage according to its overall dependencies. Experiments on the real world dataset show that our TIAO achieves highly competitive results, and that a 7B foundation model enhanced by TIAO performs comparably to GPT-4 and GPT-5-nano. Code is available at this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2609.16748
PDF
-
Protocol-Preserving Context Trimming for Agentic Workflows: Benefits, Failure Regimes, and Budget Guardrails
Abstract
Agentic large language model (LLM) systems rely on long interaction histories to preserve instructions, tool states, intermediate decisions, and unresolved dependencies, but unrestricted context growth increases computational cost and can reduce efficiency. This study evaluates protocol-preserving context trimming as a reliability-constrained approach for multi-step agentic workflows. Five trimming strategies - recency-based, relevance-based, summarization, protocol-aware trimming, and adaptive budget guardrails - were compared across retained-context levels and workflow-complexity classes using task success, protocol adherence, valid tool calls, token savings, latency reduction, cascading failures, and critical context thresholds. Conventional strategies achieved about 60% mean token savings but lower task success (66.6-77.3%) and protocol adherence (85.5-88.6%). Protocol-aware trimming improved task success to 92.2%, while adaptive guardrails achieved 96.0% task success, 96.3% protocol adherence, and 1.0% cascading failure with 56.0% mean token savings. Retained-context budgets of 25% or less increased failure odds 10.92-fold relative to budgets of 50% or more (p < 0.001). Protocol-aware trimming produced 5.24-fold greater odds of successful completion than conventional methods under aggressive budgets, while adaptive guardrails further increased success odds 2.11-fold versus fixed protocol-aware trimming (p < 0.001). Critical context thresholds also increased with workflow complexity. These findings indicate that reliable context reduction depends more on preserving protocol-critical state than on maximizing token removal, and that adaptive guardrails can improve efficiency, scalability, and reliability in long-horizon agentic systems.
Abstract (translated)
URL
https://arxiv.org/abs/2609.16461
PDF
-
KnowBench: Effort Reduction as a Unified, Deployment-Grounded Benchmark for Clinical AI
Abstract
Clinical AI systems are evaluated with instruments built for research settings (reference-based similarity metrics and expert rubric panels) that measure resemblance to an artifact rather than reduction of a burden. We introduce KnowBench, pioneered by Knowtex, whose unifying metric is Effort Reduction (ER): the proportion of system-generated clinical work product accepted by the responsible clinician under expert and safety review. ER is defined once and instantiated per task across the administrative workload clinical AI automates: visit notes, diagnosis and billing codes, orders, EHR chart summarization, patient after-visit summaries, and clinical decision support. In every instantiation the construction is identical: the clinician's review-and-attestation event is the ground truth, every accepted unit is work the system completed, and every correction is residual effort returned to the clinician. The primary contribution of this paper is the benchmark itself: the metric, its degenerate cases, and a reporting protocol under which ER claims are auditable and cross-system comparable. Alongside it we report an initial headline measurement from the documentation instantiation: over one million signed encounters across a production window exceeding six months and thirteen medical specialties, Knowtex's proprietary fine-tuned clinical foundation models operating inside a closed feedback architecture achieve an aggregate ER of 97.99%, with per-specialty aggregates spanning 96.8-98.9%. This release reports the protocol's checklist partially, and states which companion statistics are withheld; the benchmark is offered so that this figure, and every figure reported after it, can be held to the same standard.
Abstract (translated)
URL
https://arxiv.org/abs/2609.15794
PDF
-
New Conditions for Philosophers to Catch the Wave of Citizen Deliberation in the Age of Artificial Intelligence in advance
2026-09-14 15:11:02Bernard Reber (CEVIPOF)Abstract
Powerful technologies labeled ``AI''-without sufficient epistemic caution-are already reshaping political and private life, bringing both new dangers and new opportunities for citizen participation. These range from electoral and legislative engagement to the most ambitious form: political co-creation through citizens' assemblies. Large Language Models (LLMs) could support such processes through moderation, translation, facilitation, summarization, and writing assistance. But this potential remains largely unrealized. The Democratic Commons project takes a fundamentally interdisciplinary approach-from philosophy to computer science-to evaluate LLMs against five proposed democratic principles. At its core, the project is driven by the question of political bias: under what conditions can LLMs be used democratically within forms of citizen participation that are them- selves still largely experimental? Addressing these socio-technical questions requires grounding in political theory and, more broadly, in philosophy-disciplines that provide the normative frameworks without which the democratic evaluation of AI systems cannot be mean- ingfully conducted.
Abstract (translated)
URL
https://arxiv.org/abs/2609.15707
PDF
-
DA-DLM: Explicitly Modeling Token Dependencies in Diffusion Language Models
2026-09-14 05:39:12Pengyu Ji, Zichen Zhang, Xiang Hu, Kewei TuAbstract
Diffusion Language Models (DLMs) generate text by iteratively denoising a masked sequence, independently predicting multiple tokens at each step. This conditional independence discards inter-token dependencies and degrades coherence-an issue that parallels the multi-modality problem in Non-Autoregressive Translation (NAT). Drawing on the Directed Acyclic Transformer (DAT), which tackles this problem in NAT via a Directed Acyclic Graph (DAG), we propose DA-DLM, a model that adapts DAG-based dependency modeling to DLMs' iterative setting through a position-oriented DAG design. The position-oriented DAG binds node groups to fixed output positions so that tokens fixed in earlier steps anchor neighboring predictions via learned transitions, and evolves with denoising to focus on remaining uncertainty as anchors accumulate. On language modeling, open-ended generation, and summarization, DA-DLM consistently outperforms Block Diffusion, especially under fewer denoising steps, and matches autoregressive models while preserving the parallel generation advantage. Our code is publicly available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2609.15070
PDF
-
Zero-shot video highlight detection based on text descriptions and synthetic images
2026-09-13 21:03:57Michal Byra, Alberto Presta, Grzegorz Stefanski, Krzysztof ArendtAbstract
Detecting video highlights, the most informative or engaging moments in a video, is important for applications such as video summarization and content recommendation. We propose a zero-shot framework that combines CLIP, large language models (LLMs), and diffusion models. Given lightweight video metadata, such as a title or category, an LLM generates textual descriptions of likely highlight events. These descriptions are further converted into synthetic visual prototypes using a diffusion model. Textual and visual representations are matched to video frames using CLIP, enabling frame-level highlight detection without highlight annotations or dataset-specific training. Experiments on TVSum and SumMe demonstrate strong zero-shot performance, with particularly favorable results on TVSum. The proposed approach provides an effective framework for metadata-conditioned zero-shot video highlight detection.
Abstract (translated)
URL
https://arxiv.org/abs/2609.14790
PDF
-
Pull: Lazy Materialization of Working Memory for Stateful LLM Conversations
Abstract
As LLM conversations grow to hundreds of turns, full-context injection incurs $O(N^2)$ cumulative token costs, while lossy summarization or hard truncation irreversibly discards historical state. We propose Pull, a session router that maintains an addressable metadata directory via a local, deterministic Purifier (zero LLM calls, millisecond-level latency). At query time, the LLM lazily materializes only the turns it needs; unmaterialized turns remain accessible but collapsed. Unlike irreversible compression, Pull's materialization is reversible; subsequent queries can expand any collapsed turn. On LoCoEval (128 conversations, 12,780 turns), Pull reduces per-query context tokens (Phase 2) by 75.1 percent on single-hop tasks with equivalent quality ($\Delta = -0.002$, n.s.) and by 72.0 percent on multi-hop tasks with no quality loss ($\Delta = +0.017$). A controlled routing benchmark (7,831 queries x 10 methods) shows that entity lifecycle tracking is empirically a prerequisite for distance-independent routing. On BEAM 1M (14 conversations, 263 questions), Pull improves F1 by +55.2 percent over a truncation baseline.
Abstract (translated)
URL
https://arxiv.org/abs/2609.14773
PDF
