<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="das-2026-time">
    <titleInfo>
        <title>TIME: Temporally Intelligent Meta-reasoning Engine for Context-Triggered Explicit Reasoning</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Susmit</namePart>
        <namePart type="family">Das</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2026-07</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Findings of the Association for Computational Linguistics: ACL 2026</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Maria</namePart>
            <namePart type="family">Liakata</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Viviane</namePart>
            <namePart type="given">P</namePart>
            <namePart type="family">Moreira</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jiajun</namePart>
            <namePart type="family">Zhang</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">David</namePart>
            <namePart type="family">Jurgens</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">San Diego, California, United States</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
        <identifier type="isbn">979-8-89176-395-1</identifier>
    </relatedItem>
    <abstract>Reasoning-oriented language models typically expose explicit reasoning as a long, front-loaded chain of “thinking” tokens before the main output, either always enabled or externally toggled at inference time. Although this can help on arithmetic, coding, and other multi-step tasks, it is costly, weakens claim-level auditability, and does not allow the model to re-trigger explicit reasoning once presentation has begun. In dialogue, these limitations are compounded by weak sensitivity to temporal structure: unless time is explicitly stated in text, standard models treat replies separated by seconds and replies separated by weeks as equivalent. We introduce TIME (Temporally Intelligent Meta-reasoning Engine), a behavioral alignment framework that learns explicit reasoning as a context-triggered control policy rather than a fixed response mode. TIME augments dialogue with optional ISO 8601 ‘\ensuremath&lt;time\ensuremath&gt;‘ tags, tick events that represent silent time passage, and short ‘\ensuremath&lt;think\ensuremath&gt;‘ blocks that may appear anywhere in a response. Using a four-phase curriculum, including a small maximally diverse full-batch alignment stage, we train Qwen3 dense models to invoke brief, in-place reasoning bursts only when contextual cues warrant them, while keeping user-facing output compact. We also introduce TIMEBench, a diagnostic benchmark for evaluating reasoning from temporal cues in dialogue. Across 4B-32B scales, TIME improves TIMEBench scores over the corresponding base Qwen3 models in both thinking and no-thinking modes while reducing explicit reasoning tokens by roughly an order of magnitude. Beyond score improvements, TIME induces a distinct behavioral shift: explicit reasoning becomes more compact and more responsive to contextual cues. Code, training data, and benchmark artifacts are publicly available.</abstract>
    <identifier type="citekey">das-2026-time</identifier>
    <identifier type="doi">10.18653/v1/2026.findings-acl.1936</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-acl.1936/</url>
    </location>
    <part>
        <date>2026-07</date>
        <extent unit="page">
            <start>38872</start>
            <end>38906</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T TIME: Temporally Intelligent Meta-reasoning Engine for Context-Triggered Explicit Reasoning
%A Das, Susmit
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F das-2026-time
%X Reasoning-oriented language models typically expose explicit reasoning as a long, front-loaded chain of “thinking” tokens before the main output, either always enabled or externally toggled at inference time. Although this can help on arithmetic, coding, and other multi-step tasks, it is costly, weakens claim-level auditability, and does not allow the model to re-trigger explicit reasoning once presentation has begun. In dialogue, these limitations are compounded by weak sensitivity to temporal structure: unless time is explicitly stated in text, standard models treat replies separated by seconds and replies separated by weeks as equivalent. We introduce TIME (Temporally Intelligent Meta-reasoning Engine), a behavioral alignment framework that learns explicit reasoning as a context-triggered control policy rather than a fixed response mode. TIME augments dialogue with optional ISO 8601 ‘\ensuremath<time\ensuremath>‘ tags, tick events that represent silent time passage, and short ‘\ensuremath<think\ensuremath>‘ blocks that may appear anywhere in a response. Using a four-phase curriculum, including a small maximally diverse full-batch alignment stage, we train Qwen3 dense models to invoke brief, in-place reasoning bursts only when contextual cues warrant them, while keeping user-facing output compact. We also introduce TIMEBench, a diagnostic benchmark for evaluating reasoning from temporal cues in dialogue. Across 4B-32B scales, TIME improves TIMEBench scores over the corresponding base Qwen3 models in both thinking and no-thinking modes while reducing explicit reasoning tokens by roughly an order of magnitude. Beyond score improvements, TIME induces a distinct behavioral shift: explicit reasoning becomes more compact and more responsive to contextual cues. Code, training data, and benchmark artifacts are publicly available.
%R 10.18653/v1/2026.findings-acl.1936
%U https://aclanthology.org/2026.findings-acl.1936/
%U https://doi.org/10.18653/v1/2026.findings-acl.1936
%P 38872-38906

Markdown (Informal)

[TIME: Temporally Intelligent Meta-reasoning Engine for Context-Triggered Explicit Reasoning](https://aclanthology.org/2026.findings-acl.1936/) (Das, Findings 2026)

• TIME: Temporally Intelligent Meta-reasoning Engine for Context-Triggered Explicit Reasoning (Das, Findings 2026)

ACL

• Susmit Das. 2026. TIME: Temporally Intelligent Meta-reasoning Engine for Context-Triggered Explicit Reasoning. In Findings of the Association for Computational Linguistics: ACL 2026, pages 38872–38906, San Diego, California, United States. Association for Computational Linguistics.