Computation and Language
CertMark: Distortion-Free Multi-Bit Watermarking with Certified Decoding
Paweł Batorski, Przemysław Spurek and Paul Swoboda
Leading multi-bit watermarking methods for
language models encode messages by biasing the model's next-
token probabilities, creating a trade-off between message recovery and text quality. Their decoders typically return the highest-scoring candidate from accumulated
token-level evidence, without a certified abstention rule that bounds the probability of outputting an incorrect message. We introduce CertMark, a distribution-preserving multi-bit watermark with certified decoding. Rather than modifying probabilities, CertMark uses the embedded message to seed an exact Gumbel-max sampler, thereby preserving the model's original sampling distribution. We propose two scalable decoders: a model-agnostic, text-only decoder and a model-aware variant that leverages the original next-
token distributions for stronger recovery. Both support certified abstention with mathematical bounds on the probability of returning an incorrect message. Across text completion, summarization, and story generation, CertMark matches the perplexity of unwatermarked text while reliably recovering multi-bit messages. The model-aware decoder further achieves higher bit accuracy than probability-biasing baselines. Our code is publicly available at https://github.com/Batorskq/CertMark.
Read the full abstract ▾
Computation and Language · Cross-listed from cs.DB
Robust Hierarchical Structures for Agentic Document Analysis
Ruiying Ma, Yiming Lin and Aditya G. Parameswaran
Large
Language Models (LLMs) enable us to better understand text documents, including PDFs and Word documents. However, LLMs, as well as more modern LLM
agents, i.e., those with tool-calling abilities, typically treat such documents as plain text, ignoring the fact that they are often organized hierarchically into sections and subsections. Extracting this structure, while difficult, can improve efficiency and effectiveness for
agents (and humans)—since only sections relevant to a given task need to be processed. Unfortunately, prior work on structure extraction provides no formal guarantees on how well the inferred structure matches the true one. Instead, we target a robust and compact variant that is feasible to infer and useful in practice. Robustness ensures that the text under each subsection header is a superset of the text under the same header in the true structure. Compactness seeks to minimize this superset, reducing agentic cost (or human cognitive load). We propose SHED, a two-stage workflow for inferring a robust and compact structure. The first stage is pluggable with an infinite family of approaches, each guaranteeing robustness for a specific document class. We theoretically characterize the document space using these classes and their hierarchical relationships. Empirically, SHED improves F-1 scores (measuring the robustness–compactness trade-off) by 13%–68% over non-LLM baselines and 9%–15% over expensive LLM-based approaches. Finally, we show how SHED-inferred structures are valuable for agentic document analysis:
agents using SHED outperform baselines, achieving 3%–23% higher accuracy while being up to 10x cheaper.
Read the full abstract ▾
Computation and Language
Jianguo Huang, Lipeng Wan, Yanchen Deng et al. (+1 more)
Diffusion large
language models (dLLMs) generate text by iterative unmasking. At each denoising step, a dLLM proposes a
token at every masked position, but the decoder commits only a confident subset of these proposals. Trace-based on-policy
distillation (TOPD) builds on this process by matching the student to a stronger teacher, yet only at the committed positions. We argue that this discards much of the useful signal, which resides in the uncommitted proposals, where the student has made a prediction but is not yet confident enough to commit it. We call these proposals hesitations. In our pilot study on an SDAR-4B student, hesitations make up only 24% of supervisable state-position pairs but carry 66% of the teacher-student divergence. To exploit this signal, we propose Hesitation-Aware On-Policy
Distillation (HOPD), which extends teacher distribution matching to every masked position of each denoising step. Because hesitations are not equally informative, we further allocate supervision using hindsight from the completed trajectory, placing more weight on positions whose proposal was later disagreed with the final
token and on blocks where first-step proposals rarely survive. Since both models already produce distributions at all masked positions, HOPD requires no additional forward passes over TOPD. The only extra cost is evaluating the loss at more positions. With SDAR-1.7B and SDAR-4B students distilled from TraDo-8B-Instruct, HOPD achieves the best average score among the evaluated methods on five math and coding
benchmarks, under both static and dynamic decoding and at both scales. It also speeds up decoding. On SDAR-4B, the HOPD student hesitates less and commits 11% more
tokens per denoising step than TOPD, while reaching higher accuracy.
Read the full abstract ▾
All authors Jianguo Huang, Lipeng Wan, Yanchen Deng, Bo An
Computation and Language
BaatCheet: A Multilingual Corpus for Dialogue Translation in Indian Languages
Priyanka Dasari, Yuvrajsinh D. Bodana, Vandan Mujadia et al. (+3 more)
Existing translation models are typically trained on sentence-level and formal text, limiting their ability to capture everyday conversational dialogue phenomena such as informality, speaker interaction, and discourse coherence. Most existing Indic translation resources and evaluation
benchmarks focus on sentence-level or formal text, making it difficult to assess translation quality of the dialogue phenomena. In this work, we introduce BaatCheet, a multilingual dialogue corpus named after the Hindi term for conversation or chitchat, containing approximately 49,000 dialogues for dialogue translation across five translation directions. We fine-tune five open-source LLMs across seven training data configurations and find that fine-tuning yields substantial gains over zero- and few-shot baselines. To comprehensively evaluate dialogue translation quality, we employ multiple evaluation strategies, including automatic metrics, LLM-as-judge, and human assessments using an SQM-guided Direct Assessment (DA) Protocol.
Read the full abstract ▾
All authors Priyanka Dasari, Yuvrajsinh D. Bodana, Vandan Mujadia, Arafat Ahsan, Dipti Misra Sharma, Parameswari Krishnamurthy
Computation and Language · Cross-listed from cs.AI
TraceDance: An Automated System for Building Agent Behavior Benchmarks from Real-World Agent Deployment Traces
Dehai Min, Daoan Zhang, Yiming Zeng et al. (+13 more)
An
agent can complete a task while exhibiting undesirable behavior during execution. Developers need tests for the specific behaviors encountered in deployment, beyond fixed
benchmark suites. We present TraceDance, an
agent system that constructs targeted
benchmarks from deployment traces for user-specified undesirable behaviors. For efficient construction, Anchor-and-Confirm combines programmable
retrieval with candidate-level confirmation by a Flash large
language model (LLM), while the Anchor Synthesis Loop generates and revises specifications for custom behaviors. The
benchmarks use decision-point continuation to evaluate an LLM's next turn at a recorded decision point with a behavior-specific rubric, without a reference answer or environment replay. Experiments in coding and general tool use draw on 252,557 sessions and produce 107
benchmarks with 4,125 instances, fulfilling 95.3% of build-target requests. Both human annotators confirm the requested behavior in 84% of sampled instances, and the automated grader's agreement with human pass/fail judgments is comparable to that between the annotators. Nine frontier LLMs achieve a mean pass rate of only 26.7%, showing that they still struggle to respond appropriately at the evaluated decision points. Analysis across behavior-specific
benchmarks further reveals weaknesses in how current LLMs behave as
agents. By turning deployment problems into targeted
benchmarks, TraceDance could serve as a key component of the recursive self-improvement (RSI) loop.
Read the full abstract ▾
All authors Dehai Min, Daoan Zhang, Yiming Zeng, Huayi Zhang, Ziyi Chen, Yan Zhang, Qinbo Bai, Mengyuan Chao, Jing Ning, Qiyue Hua, Huiyi Chen, Hanrong Zhang, Henry Peng Zou, Jie Yang, Wei Xu, Philip S. Yu
Computation and Language
Calibration, Not Answer Selection: Distilling Internal Confidence in Reasoning Models
Yadong Xi, Rongsheng Zhang, Tangjie Lv et al. (+2 more)
Reinforcement learning with binary correctness rewards trains correctness, not calibrated confidence. The confidence that
reasoning models verbalize is systematically overconfident, and the problem is not merely one of scale: verbalized confidence tracks how willing a model is to commit to an answer, not how likely the answer is to be right. Post-hoc rescaling therefore fits one distribution but rarely transfers. We look inside the model instead. On factual question answering, a linear probe on the hidden state between the chain of thought and the answer is substantially better calibrated: its expected calibration error is 5 to 38 times lower than that of the verbalized score across four
benchmarks and two model families. However, when used to pick among N sampled answers, that same probe nearly ties majority voting yet falls far short of the oracle. Internal states answer "how certain am I" well and "which answer is right" poorly, so the signal should be reported as a confidence rather than used to select answers. As a result, we introduce probe-guided self-
distillation (Probe-SD): score a model's own sampled traces with the probe, overwrite the confidence each trace states, and finetune the base checkpoint of the same family, so nothing but the model itself remains at test time. On Qwen3-14B, Probe-SD cuts ECE from 0.178 to 0.024 in-domain and from 0.542 to 0.113 out-of-domain, where it also beats post-hoc recalibration and self-consistency
distillation. The resulting confidence is well-calibrated and useful for weighted voting, behaviors previously attributed to online RL, here obtained with supervised finetuning alone.
Read the full abstract ▾
All authors Yadong Xi, Rongsheng Zhang, Tangjie Lv, Ziyang Luo, Ruochen Zhao
Computation and Language · Cross-listed from cs.CV
InfoEdit: Probing Global Layout Reasoning in Infographic Editing
Cheng Yang, Chufan Shi, Huijuan Wang et al. (+6 more)
Multimodal foundation models edit natural photographs at production quality, yet the same models struggle with structured visual content such as infographics. Unlike photographs, infographics encode information through logical relations; editing one element often requires surrounding elements to be adapted. We refer to this global layout
reasoning capability as reflow. Existing image-editing
benchmarks neither provide a dedicated setting for structured visual content nor evaluate the reflow capability. We introduce InfoEdit, a novel
benchmark of 1,000 infographics across eight logical-relation families, paired with 4,000 editing instructions across four editing tasks, and a reflow-aware evaluation protocol. Across eight frontier editors, only GPT-Image-2 clears 60% average success rate; most models fall below 7%, and no editor exceeds 36% on the Swap-Block task even with perfect target localization. We further show that code-level editing can match the strongest pixel-level editor, revealing complementary strengths across tasks. InfoEdit identifies reflow as a central challenge in structured visual content editing and provides a diagnostic
benchmark to facilitate future progress.
Read the full abstract ▾
All authors Cheng Yang, Chufan Shi, Huijuan Wang, Bo Shui, Yaokang Wu, Muzi Tao, Yibo Yan, Xuezhe Ma, Taylor Berg-Kirkpatrick
Computation and Language
The Text Beside the Image: Detection, Utility and Leakage for Trustworthy Multimodal Medical Data and Beyond
Andreas Maier, Monica Hinrichs-Mayer, Franziska Weber et al. (+4 more)
Medical images are released with the reports that describe them, and protecting the image does not protect the report. This paper measures the text component of such releases. We measure identifier detection, downstream utility and residual identity leakage on the same documents, with the pseudonymisation policy as the variable under test: 15 detectors, three release conditions and four corpora of medical reports, legal judgments, news and other genres, and e-mail, in German, English, Chinese and Arabic. A fixed 13-detector union reaches a person sensitivity of 0.9998 at specificity 0.8686 on the medical reports, 0.9958 at 0.8504 on the legal judgments, 0.9352 at 0.9318 on news and other genres, and 0.9906 at 0.6235 on e-mail. With this ensemble, frequency matching with a public name list recovers zero identities by
alignment across the four corpora; the names it got right were ones the detector missed, left in clear text. Cross-document linkage ranks the correct person first for 0.93% of e-mail queries without training and 3.94% with it, against 1/3697 chance and 71.98% on unmodified text. On the medical reports it recovers nothing without training and 0.71% of 138 queries with it, against 1/207 chance and a 2.73% ceiling on unmodified text.
Read the full abstract ▾
All authors Andreas Maier, Monica Hinrichs-Mayer, Franziska Weber, Niklas Lackner, Matthias May, Bernhard Kainz, Siming Bayer
Computation and Language · Cross-listed from cs.LG
BERT4DTI : BERT-based Model for Predicting Drug-Protein Interactions
Thanina Hamitouch, Khadidja Henni, Abdelkrim Arie et al. (+3 more)
Understanding how drugs interact with
protein targets is fundamental to drug discovery, drug repurposing and the early identification of promising therapeutic candidates before costly experimental testing. Sequence-based DTI models face three practical limitations: labelled interactions are scarce and unevenly distributed, large pretrained chemical and
protein encoders are expensive to fine-tune end-to-end, and independently encoded sequences do not capture pair-specific dependencies. We present BERT4DTI, which encodes SMILES strings with ChemBERTa and amino-acid sequences with ProtBERT, applies bidirectional mutual
attention between
token-level representations, and classifies the resulting interaction features using convolutional layers and a multilayer perceptron. To reduce trainable size, ProtBERT is truncated to 18 retained layers and only the last two layers of each encoder are fine-tuned. On BIOSNAP, DAVIS and BindingDB, BERT4DTI is competitive, achieving the best ROC-AUC and PR-AUC on BIOSNAP and the highest sensitivity on all three
benchmarks. An ablation on DAVIS shows that mutual
attention improves PR-AUC and specificity. With 125M trainable parameters compared with 353M for full BERT fine-tuning, BERT4DTI provides a favourable performance-parameter trade-off for sequence-based DTI screening, while leaving runtime profiling, calibration and leakage-audited validation for future work.
Read the full abstract ▾
All authors Thanina Hamitouch, Khadidja Henni, Abdelkrim Arie, Amina Selma Haichour, Neila Mezghani, Lina Abou-Abbas
Computation and Language · Cross-listed from eess.AS
Acoustic Progress Propagation for Long-Horizon Speculative Decoding in ASR
Yuanyuan Jia and Qianqian Yang
Speculative decoding accelerates autoregressive automatic
speech recognition (ASR), but the acceptance length of
alignment-aware drafters can saturate as the draft horizon increases. We propose a progress-aware speculative drafter that recurrently propagates an acoustic progress state across draft steps and feeds it back into audio cross-
attention to guide
token generation. We jointly train the drafter and progress predictor over variable draft horizons. On five ASR test sets, our method achieves lossless, macro-averaged end-to-end speedups of 1.657x and 1.227x over target-only autoregressive decoding with Qwen3-ASR-0.6B and Qwen3-ASR-1.7B, respectively. Relative to AnchorDraft, our method improves the macro-averaged speedup by 34.3% and 9.0%, respectively. Horizon sweeps show continued growth in acceptance length beyond the baselines' saturation. Code is available at https://github.com/yuanyuanjia71-spec/ProgDraft.
Read the full abstract ▾
Computation and Language
Beyond Memory Construction: Rethinking Memory Access for LLM-based Conversational Agents
Donghua Cai, Yongheng Deng, Yifei Wang et al. (+2 more)
Memory is a core component of conversational
agents, enabling coherent and context-aware behavior over long interactions. Recent approaches commonly rely on LLM-based
memory construction, where raw interactions are rewritten into structured
memory units and later retrieved via a RAG pipeline. While effective in controlled settings, we show that this paradigm breaks down in long-horizon, high-
entropy conversations:
memory construction becomes increasingly lossy and unstable as context length and information complexity grow, and incurs prohibitive cost due to repeated LLM invocation. To address these limitations, we propose Threader, a
memory system that shifts the focus from
memory construction to efficient, structure-aware access over raw interactions. Instead of rewriting interactions, Threader preserves them as first-class
memory, organizes them into topic-coherent segments via lightweight incremental segmentation, and enables accurate
retrieval through multi-view representation. At query time, it performs multi-signal
retrieval that combines segment-level access with localized evidence matching, ensuring both completeness and coherence. Extensive experiments demonstrate that Threader consistently improves answer accuracy and evidence recall, while significantly reducing the
memory construction overhead.
Read the full abstract ▾
All authors Donghua Cai, Yongheng Deng, Yifei Wang, Zijun Shen, Ju Ren
Computation and Language · Cross-listed from cs.LG
When Do Models Admit They Are Wrong? Failure Disclosure Is Unstable Under Reinforcement Learning
Steven Y. Feng, Noah D. Goodman, Michael C. Frank et al. (+3 more)
Outcome-based reinforcement learning can produce models with similar task performance but very different ways of communicating about their mistakes. We study failure disclosure: whether a model admits that an attempted solution failed rather than staying silent or presenting it as successful. Across repeated outcome-only GRPO training runs, failure disclosure varies far more than task accuracy. The pattern extends to a second
reasoning task and stabilized PPO, persists at 7B, and also appears in an instruction-conditioned 32B setting. We also find that small floating-point and sampling differences during training can redirect reporting behavior even when the task objective and earlier training history are held fixed. Additional tests show that failure disclosure is not a single decision: Checking the answer, entering a report, and completing the admission can separate, and the weak point depends on the task and response format. Further, experiments with neutral controls show more broadly that behaviors left weakly constrained by training are especially likely to vary across runs, of which failure disclosure is an example. We can reduce variability in failure disclosure by discouraging the model from drifting from its starting policy on failed, well-formed responses. This makes reporting substantially more consistent, though its effect on task performance depends on the setting. Stable task accuracy therefore does not guarantee stable safety-relevant behavior: Researchers should measure these behaviors directly across runs and design training methods that keep them reliable.
Read the full abstract ▾
All authors Steven Y. Feng, Noah D. Goodman, Michael C. Frank, Evan Hubinger, Paul C. Bogdan, Andrew Lampinen
Computation and Language
CoLMbo-SV: A Grounded Language Model for Explainable Speaker Verification
Massa Baali, Sarthak Bisht, Ziyue Qiu et al. (+3 more)
Speaker verification systems achieve high accuracy but provide little account of the acoustic evidence behind their judgments. Making these systems inspectable requires exposing interpretable evidence while retaining the richer information on which their decisions depend. We present CoLMbo-SV, a speaker
language model that combines strong speaker discrimination with structured, acoustically grounded comparison reports. By connecting a pretrained speaker encoder to a
language model and supplying explicit acoustic measurements, CoLMbo-SV makes voice comparisons inspectable without restricting verification to the evidence verbalized in its reports. We additionally introduce VoxReason, paired recordings with measured acoustic properties and comparison reports filtered through numerical and qualitative checks, providing supervision for this combined capability. We also develop an evaluation framework that separates what acoustic information a speaker representation encodes, what influences the verification score, and what the generated report discusses. On VoxCeleb1-O, CoLMbo-SV achieves 0.99% EER, reducing verification error by approximately 80% relative to the strongest audio-
language baseline fine-tuned on VoxReason, while attaining a numerical-grounding score of 0.82. Our analysis further demonstrates that acoustic correctness and decision relevance are distinct properties of an explanation, exposing a gap that numerical-grounding metrics miss. Together, these contributions substantially advance audio-
language speaker verification, bring its accuracy toward that of dedicated speaker encoders while adding checkable acoustic reporting, and establish an empirical framework for connecting natural-
language explanations to the decisions they explain.
Read the full abstract ▾
All authors Massa Baali, Sarthak Bisht, Ziyue Qiu, Joseph Konan, Rita Singh, Bhiksha Raj
Computation and Language
Beyond Calibration: Do a Typed-Decision Model's Probabilities Obey the Probability Axioms?
Keyi Li, Yihao He and Quanyi Li
Typed-decision models such as TypeSafe's Jev answer a declared yes/no or multiple-choice question about a state with a probability instead of text, and their evaluations report accuracy and calibration. Neither requires that the probabilities a model gives to logically related questions fit together item by item, which is what a system that acts on those probabilities needs. We test this property, coherence, with a battery of logically linked questions that needs no labels. For 160 items from ChaosNLI and PubMedQA, each with three mutually exclusive labels, we ask whether the label is X, whether it is not X, whether it is one of the other two labels, and which label applies. On 480 negation pairs, Jev's probabilities for "the label is X" and "the label is not X" miss summing to one by 0.064 on average (95% CI 0.055 to 0.072). Qwen3.8-27B, run from its official BF16 weights, misses by 0.293 with first-
token probabilities and by 0.122 with verbalized probabilities. The gap to the first-
token readout persists on pairs where both systems give similar probabilities, without double-negation labels, and after averaging Jev's repeated calls. Jev is not coherent either: its violations are about five times its repeat noise, and it over-endorses statements about single labels, so that its three single-label probabilities sum to 1.14 on average. The two systems also fail differently. Qwen3.8-27B's first-
token readout under-endorses the complement of a label whether or not the question contains "not", rejecting both a statement and its negation in 196 of 480 pairs, and it does not become more coherent where it is more confident, whereas Jev's violations concentrate where its answer is uncertain. Because the checks need no labels, they expose biases that appear only when question forms are compared, and inconsistencies within items.
Read the full abstract ▾
Computation and Language
Turning Speech Language Models into Multilingual Listeners
Tolúlopé Ògúnrèmí, Dan Jurafsky, Chris Manning et al. (+2 more)
Speech Language Models (SLMs) that understand spoken
language questions support only a few high-resource
languages, limiting access to millions of people worldwide. This gap stems from the scarcity of multilingual
speech instruction-tuning
datasets. We present MULTISPEECHQA, a large-scale, synthetically generated and human-verified
dataset comprising 9200 hours of 10.8 million spoken question-answer pairs in 23 typologically diverse
languages. Using MULTISPEECHQA, we also introduce MULTISPEECH-BENCH, a multi-task
benchmark for evaluating SLM performance on 23
languages. We compare the performance of a cascading system to open-weight and closed SLMs on MULTISPEECH-BENCH and find that the cascading system outperforms open-weight SLMs but not all closed SLMs. We use MULTISPEECHQA to finetune Qwen 2.5-Omni, which improves its performance on our
benchmark. Our findings show that high-quality synthetic
datasets offer a cheap solution to improving the multilingual capabilities of SLMs.
Read the full abstract ▾
All authors Tolúlopé Ògúnrèmí, Dan Jurafsky, Chris Manning, Ahmet Üstün, Martijn Bartelds
Computation and Language · Cross-listed from cs.LG
Safaeid Hossain Arib, Rabeya Akter, Ismam Nur Swapnil et al. (+3 more)
On-policy self-
distillation trains
reasoning models on their own trajectories using dense
token distribution guidance from a privileged teacher with access to a verified solution. This supervision transfers what the teacher predicts without directly transferring where it attends within the preceding context. We introduce On-Policy
Attention Self-
Distillation (OPASD), which complements
token-level supervision with solution-conditioned
attention distillation. Because the privileged teacher can attend to verified solution
tokens unavailable to the student, OPASD projects teacher
attention onto student-visible positions and renormalizes the resulting distribution before
alignment. Across three model sizes and four competition-level mathematics
benchmarks, OPASD consistently outperforms
token-only OPSD, improving average accuracy by 4.98 to 8.40 percentage points. OPASD also avoids the response-length inflation and performance degradation observed with
token-only
distillation, reducing generated rollout
tokens by 73.9% and estimated model compute by 72.6% while training 1.53x faster. These results show that solution-conditioned
attention provides a complementary supervision signal that makes on-policy self-
distillation more accurate, stable, and compute-efficient.
Read the full abstract ▾
All authors Safaeid Hossain Arib, Rabeya Akter, Ismam Nur Swapnil, Md. Faiyaz Abdullah Sayeedi, Tasnim Mohiuddin, Md Mofijul Islam
Computation and Language · Cross-listed from cs.LG
AG-CoT: Verified Algorithmic Traces for LLM Program Synthesis on Clifford Circuits
Lu Wei, Yufeng Wang, Chenfeng Cao et al. (+2 more)
Scientific code generation can produce executable programs that fail to compute the intended scientific object. We study this problem in
language-model synthesis of Clifford
circuits, which prepare the stabilizer states used in
quantum error correction and admit exact classical verification. In our target-conditioned framework, each target is given as compact signed stabilizer generators, and an exact verifier checks the generated OpenQASM
circuits. We supervise models with Aaronson-Gottesman chain-of-thought (AG-CoT) traces checked by the verifier, and continue training on model generations that the verifier accepts. Across two independently trained model families (3B and 7B), AG-CoT supervision multiplies greedy-decode state-equivalence accuracy by four to six times over
circuit-only baselines, and verifier-filtered continuation training adds a further consistent gain atop both. A complementary 32B study shows that supervised models achieve near-perfect syntax and Clifford validity while the strongest direct model reaches 6.14% state equivalence per target, rising to over 10% under verifier-guided selection with multiple candidates. These results show that algorithmic trace supervision gives a large, statistically significant gain in both model families and that verifier-filtered continuation adds a further repeated gain. The persistent gap between Clifford validity and state equivalence confirms that exact verification is necessary: a
circuit can be syntactically and physically valid yet prepare the wrong
quantum state.
Read the full abstract ▾
All authors Lu Wei, Yufeng Wang, Chenfeng Cao, Lu Pang, Haibin Ling
Computation and Language
Identifying Temporal Features within Transcoders for Time Sensitive Factual Recall
Sanjay Govindan, Yang Song and Maurice Pagnucco
Large
Language Models (LLMs) suffer from temporal misalignment, often due to the contradictory nature of their training corpora. While current mitigation strategies rely on computationally expensive fine-tuning or context-heavy
retrieval augmented generation (RAG), the internal mechanisms governing time-sensitive recall remain under-explored. Unlike prior studies that identify temporal components such as
attention heads and MLP layers, we provide the first feature-level map of temporal recall by isolating individual MLP features via transcoder
circuit tracing. We identify three node categories (common temporal, common to the year, and chrono-semantic) which interact to generate a temporal filter during factual recall. By analysing Gemma 2 2B, LLaMA 3.2 1B, and Qwen3-4B, we show that these features do not follow a simple linear pipeline but represent time through a parallel and mixed syntactic-semantic interplay across layers. We additionally discover a class of higher-layer temporal components invisible to existing EAP-IG methods, establishing transcoders as a more complete lens for temporal interpretability in time-sensitive factual recall. These findings present MLP components for potential targeted interventions in time-sensitive factual recall
Read the full abstract ▾
Computation and Language · Cross-listed from cs.AI
SeOPD: Self-Evolving LLMs via Online Policy Distillation from Self-Generated Chain-of-Thought
Xiaoshu Chen, Sihang Zhou, Ke Liang et al. (+1 more)
Recent advances in online policy self-
distillation (OPSD) have demonstrated that large
language models (LLMs) can improve their capabilities by leveraging external privileged information (PI), such as manual annotations or feedback from external environments. However, obtaining accurate annotations and constructing sophisticated environments often require substantial human effort and computation, limiting the scalability of OPSD. While a few recent studies have explored self-improvement without external PI, the resulting gains remain limited. In this work, we explore whether LLMs can achieve comparable self-improvement without external PI. Our key observation is that a single LLM can support multiple
reasoning modes, such as deep-thinking and non-thinking modes, with deep thinking generating additional information during
reasoning. Based on this observation, we propose Self-Evolving Online Policy
Distillation (SeOPD), which enables LLMs to distill and internalize information generated by their own chain of thought (CoT). Specifically, it (1) generates CoT with the deep-thinking mode, (2) produces responses with the non-thinking mode, and (3) uses the generated CoT as PI to provide
token-level supervision for the non-thinking response, allowing new information inferred during
reasoning to guide the non-thinking mode and be internalized into the shared model parameters, thereby improving both non-thinking and deep-thinking capabilities. Extensive experiments across LLMs and tasks demonstrate the effectiveness of SeOPD.
Read the full abstract ▾
All authors Xiaoshu Chen, Sihang Zhou, Ke Liang, Xinwang Liu
Computation and Language
Scoring the Wrong Question: Readout Failures in Constrained-Option Evaluation
Jiaxuan Guo, Kejia Zhang, Shuo Xin et al. (+3 more)
Constrained-option scoring reads a model's probabilities for a fixed set of permitted answers, so it returns a score even when the model is about to write something else. We study a
prompt that quotes a multiple-choice item, ending in the item's own answer instruction, and asks for a forecast of whether a reader model given only a short note will answer it correctly. The model can begin answering the quoted item: on three Qwen3 checkpoints an answer letter is the most probable
token on nearly every passage, and the forecast ranks the reader's correctness indistinguishably from chance. Beyond prior work's argmax check and prefill repair, we contribute a label-free diagnostic of the declared options' support, their first-
token mass in context; a deletion test tracing the Qwen3 answer-letter start to the quoted instruction; and rendering and tokenisation checks for dropped separators and coinciding first
tokens. Prefilling an answer stem lifts the median option mass from below to above one half on all nine checkpoints scored with this
prompt and, with options and renormalisation unchanged, raises the Qwen3-averaged AUC from 0.4929 to 0.6152. lm-polygraph's default P(True)
estimator shows a related failure: it reads True where Qwen3 without thinking favours an answer letter (MMLU) or the
prompt's (A)/(B) label (TriviaQA), and its Qwen3-averaged AUROC is indistinguishable from chance on MMLU and below chance on TriviaQA. Scoring and renormalising the (A)/(B) labels after an answer stem raises it to 0.632 and 0.868, renormalising True against False in place reaches 0.660 and 0.860, and on TriviaQA both exceed all fourteen default single-answer library
estimators. The original forecast is renormalised too, yet stays indistinguishable from chance: option mass flags low support in both settings without labels, and only checking the intended target shows which score still ranks.
Read the full abstract ▾
All authors Jiaxuan Guo, Kejia Zhang, Shuo Xin, Jingxin Yang, Youran Sun, Haizhao Yang
Computation and Language
Where Do Test-Time Scaling and Training Fall Short in Individual Stance Prediction?
Yuyang Zhao, Xuan Liu, HaoYang Shang et al. (+1 more)
Test-time scaling and post-training have improved LLM performance in coding and mathematical
reasoning, but their effectiveness for individual stance prediction remains unclear. We study this question by predicting a person's stance in a new discussion from their history. We evaluate widely used test-time scaling strategies and post-training methods, such as supervised fine-tuning and reinforcement learning, and identify four failure modes across generation, selection, and learning: (1) incorrect consensus, where repeated samples agree on the wrong stance; (2) selection failure, where generation covers the observed stance but selection misses it; (3) response overfitting, where supervised fine-tuning improves imitation but harms prediction; and (4) early plateau, where reinforcement learning shows modest initial gains followed by limited further improvement. We expose these failures using STANCE-BENCH, which contains 2499 prediction tasks from 500 Hacker News users. Guided by this analysis, we explore a simple approach that combines direct scores for all candidate stances with explicit assessments of support from the individual's history. On the 781-task test set, this approach achieves 21.83 discussion-specific Macro F1 with Qwen3-8B, compared with 19.27 for direct scoring. Our results motivate evaluating candidate generation, final selection, and person-specific evidence use separately. Our data is available at https://github.com/stance-bench/Stance-Bench.
Read the full abstract ▾
All authors Yuyang Zhao, Xuan Liu, HaoYang Shang, Haojian Jin
Computation and Language
Generalization Dynamics of LM Pre-training
Jiaxin Wen, Zhengxuan Wu, Dawn Song et al. (+1 more)
People typically assume that LMs stably mature from pattern-matching parrots to generalizable intelligence during pre-training. We build a toy eval suite and show this mental model is wrong: throughout pre-training, LMs frequently and suddenly hop between parrot-like and intelligence-like computations. We call this mode-hopping. Across our suite, LMs suddenly latch onto memorized or in-context patterns instead of in-context learning, use System 1 instead of System 2 thinking, pick up what sounds true instead of what is true, fail at multi-hop persona QA, out-of-context
reasoning, and emergent misalignment – then just as suddenly revert and generalize. Mode-hopping is not explained by standard optimization dynamics: it is locally stable and cannot be fixed by checkpoint averaging. We instead think of it as a capacity allocation problem: in a capacity-bounded model, generalizable
circuits must compete with the shallow ones learned early in training, and the data in each pre-training window may decide which
circuits win. Our suite provides a new efficient lens on generalization. We demonstrate two concrete applications: (i) select intermediate pre-training checkpoints that strongly generalize
reasoning and
alignment, better than the final pre- or mid-training checkpoints, and (ii) select pre-training data that controls and stabilizes generalization dynamics.
Read the full abstract ▾
All authors Jiaxin Wen, Zhengxuan Wu, Dawn Song, Lijie Chen
Computation and Language
CAME: Company-Aware Evidence-Memory Experts for Interpretable Quarter-Ahead Revenue Forecasting
Ya-Wen Wu, Meng-Fen Chiang, Kuang-Da Wang et al. (+1 more)
Quarter-ahead revenue forecasting requires company-scale numerical accuracy, strict temporal validity, and company-specific interpretation of narrative disclosures. LLMs can distill textual evidence but can produce scale-misaligned forecasts, whereas history-based anchors are stable but miss forecast-time signals such as product transitions, supply constraints, and management guidance. We introduce CAME (Company-Aware Evidence-
Memory Experts), a residual-forecasting framework that refines a no-leakage statistical anchor when current semantic evidence and prior error patterns justify an adjustment. On a development-inclusive rolling backtest of 336 company-quarters from 12 large public technology and platform firms, CAME achieves the lowest aggregate point-estimate error among the reported methods, with statistically supported macro-sMAPE gains over the matched Statistical Anchor, and outperforms History + Guidance on all six aggregate metrics. CAME also links adjustments to source-linked evidence cards and guarded
memory traces, supporting forecast inspection, provenance, and failure localization.
Read the full abstract ▾
All authors Ya-Wen Wu, Meng-Fen Chiang, Kuang-Da Wang, Wen-Chih Peng
Computation and Language
Knowing Is Not Choosing: What Explicit Verification Adds Beyond Generative Preference
Yilong Li, Chengpo Yan, Aayan Arish et al. (+1 more)
Generating a correct answer does not mean that a
language model will select it. We separate factual recall into three steps: generating a correct candidate, ranking the available candidates, and selecting the final answer. Pre-generation readouts predict factual recall and which questions sampling will cover across three model families, but say little about whether an available correct answer will ultimately be selected. Explicit verification with $P(\mathrm{True})$ improves within-question ranking over mean log-likelihood in Gemma, Qwen3, and Llama, with AUROC gains of $0.08$–$0.12$. In a prospectively defined Gemma cohort, verification raises plurality accuracy by about $5$ points, and still gains about $2$ points over chat-template likelihood, a stronger generative baseline. The advantage is strongest for relations with common-answer priors and depends on access to the entity; masking the entity removes the ranking advantage in larger Qwen models. Finally, the measured benefit depends on how correctness is defined: recall-oriented reference matching can credit option lists favored by likelihood and substantially understate the improvement seen under human semantic judgments. Prior work shows that models can carry latent factual knowledge and judge candidate answers; we show that these capabilities do not collapse into a single notion of “knowing,” and trace where information is gained, lost, or mismeasured between availability, ranking, and final choice.
Read the full abstract ▾
All authors Yilong Li, Chengpo Yan, Aayan Arish, Suman Banerjee
Computation and Language · Cross-listed from cs.LG
Save Your Saturated Data: Learning Beyond Reward Saturation in Group-Based RL
Ziyuan Yang, Yike Wang, Shangbin Feng et al. (+1 more)
Group-relative reinforcement learning (RL) relies on reward variation among sampled responses to estimate informative relative advantages. As
language models become increasingly capable, existing training data can become reward-saturated: all sampled responses to the same problem might receive equally high rewards, where the group-relative learning signals vanish and leave previously useful data obsolete. In this work, we investigate whether useful learning signals can be recovered from such saturated data. We study interventions at four levels of group-policy RL pipelines—data, rollout, reward, and advantage—and conduct extensive RL training on saturated
reasoning data only. While standard GRPO on saturated data would almost always yield near-0 advantages and near-noise signals, diverse interventions successfully recycle and repurpose such data: among the proposed strategies, interventions at rollout generation are consistently most effective: nudging the policy to generate “high-quality”, incorrect solutions introduces rollouts with poor rewards into saturated groups as negative samples, which turns out to improve GRPO by 6.4% to 9.0% across Qwen3-1.7B and 4B. Other interventions such as increasing rollout temperature or adding auxiliary rewards can also restore non-zero advantages, but yield less consistent gains. Further analyses show that effective negative rollouts require informative negative trajectories, that the method remains effective alongside unsaturated data, and that it supports iterative recycling of newly saturated examples. While increasingly stronger LLMs would render more data as saturated, our results demonstrate that don't waste your saturated data: with the right strategies they can be recycled into useful RL training signals in an increasingly data-scarce world.
Read the full abstract ▾
All authors Ziyuan Yang, Yike Wang, Shangbin Feng, Yulia Tsvetkov
Computation and Language
Classifying Dominant Temporal Orientation without Pretrained Text Embeddings: A Novel Morphosyntactic Inventory Vector Approach
Jonathan Cleveland and Peter S. Bearman
Computational methods have consistently struggled to determine the dominant temporal orientation of a sentence. This difficulty is especially pronounced when a sentence contains multiple embedded clauses with competing tense and aspectual information. To address this difficulty, we propose an alternative approach for identifying a sentence's past, present, or future global reference interval. Our method does not use any form of pretrained
embeddings. We rather encode sentences using fixed-length inventory vectors that are comprised of part-of-
speech counts, dependency relation counts, and explicit futurate pattern counts. We term this inventory vector of a sentence a "Morphosyntacton". The method does not use any padding, sequence models, or large
language models. Evaluation on 1,799 syntactically complex English sentences, annotated as past, present or future, shows balanced and high accuracy multiclass classification, achieving an overall multi-class accuracy of 92%.
Read the full abstract ▾
Computation and Language
MedRouter: Demystifying Knowledge Differences Across Medical LLMs for Routing-Based Reasoning
Lang Cao, Binghang Lu, Yuhao Shen et al. (+1 more)
Medical question answering spans diverse specialties and modalities, and individual medical large
language models (LLMs) exhibit distinct strengths across tasks and domains. This heterogeneity suggests that combining specialists may enable broader coverage of medical questions than relying on any single model. However, existing LLM routing methods primarily seek to balance answer quality and inference cost, leaving open how to exploit differences in specialist competence to improve medical
reasoning. In this paper, we introduce MedRouter, an agentic system that uses an
embedding-based multi-label router to select and query specialist LLMs, then passes their responses to a generator to produce the final answer. We further propose SCALE (Specialist Competence-Aware Learning), a two-stage training framework that first trains the Router with specialist correctness supervision and then optimizes its selections through reinforcement learning. The second stage uses a Performance Gain Reward (PGR) that measures how specialist information affects the generator's answer correctness relative to answering without that information. Experiments on eight text-based and
multimodal medical QA
benchmarks show that MedRouter outperforms the strongest routing baseline by 8% in average accuracy. Our analysis of specialist outputs further reveals distinct strengths and complementary question-level coverage, motivating learned routing to combine these capabilities for more comprehensive medical
reasoning.
Read the full abstract ▾
All authors Lang Cao, Binghang Lu, Yuhao Shen, Yue Guo
Computation and Language · Cross-listed from cs.LG
ECG-Scroll: A Long-Horizon, Streaming Benchmark and Agent Environment for Interpretation of Ambulatory Electrocardiograms
Haitao Li, Chenglin Li, Zhengyao Ding et al. (+3 more)
Multimodal large
language models (MLLMs) can now interpret a standard ten-second, twelve-lead electrocardiogram (ECG) with clinically grounded, reward-verified
reasoning. Real cardiac monitoring is different. Ambulatory (Holter) and telemetry recordings span hours to days and are read as they stream in, and their clinically decisive findings are paroxysmal, brief episodes buried in an otherwise unremarkable trace. Such a recording cannot be held in one context at diagnostic resolution, and its future has not yet happened, so a reader must work online, deciding what to measure now, committing evidence to
memory as it passes, and reporting events as they occur. We recast long-duration ECG interpretation as a long-horizon, online (streaming, causal) sequential decision process and introduce ECG-Scroll. As a
benchmark, long ambulatory recordings are streamed to an
agent chunk by chunk, and it must localize, quantify, and promptly flag paroxysmal events without access to future signal; because the underlying signal is retained, every answer is checkable against objective ground truth, giving rule-based rather than judge-based rewards, and the streaming formulation adds a metric batch evaluation cannot express, the detection latency between an event's onset and the moment the
agent records it. As an
agent environment, it is a fixed, gym-style interaction layer that exercises three competencies single-glance ECG models never touch:
Memory, Tool use through signal-grounded measurement rather than reading pixels, and Planning of what to measure now and when to commit. We release 390 whole-recording instances spanning 2,536 hours of two-lead ambulatory ECG and evaluate a signal-threshold rule
agent alongside off-the-shelf LLM
agents online, characterizing how they use
memory, tools, and planning and where the
benchmark's head-room lies.
Read the full abstract ▾
All authors Haitao Li, Chenglin Li, Zhengyao Ding, Ziyu Li, Yiheng Mao, Zhengxing Huang
Computation and Language
OneSign: Unifying Sign Language Understanding Tasks with One Model
Shiwei Gan, Yafeng Yin, Xiao Liu et al. (+3 more)
SLU encompasses a diverse set of tasks, including ISLR, CSLR, and SLT. Although these tasks share basic semantic and linguistic foundations, they are typically addressed with task-specific architectures and training pipelines, which hinders knowledge sharing and requires costly pretraining and finetuning for each task. In this paper, we focus on two aspects of SLU tasks: (1) training and inference pipelines are highly fragmented: most methods rely on pretraining on large-scale SL
datasets followed by task- or
dataset-specific finetuning, which leads to multiple specialized models rather than a single checkpoint. (2) current LLM-based methods may overlook the inherent modality discrepancy between sign and text
tokens, simply concatenating them and processing both modalities with the same decoder layers. In this paper, we present OneSign, a unified framework that addresses multiple SLU tasks within a single model and a single checkpoint. OneSign reformulates ISLR, CSLR, and SLT under a single training paradigm. To accommodate the heterogeneous characteristics of sign and text representations, we introduce a Modality-Adaptive Mixture-of-Experts (MA-MoE) architecture, consisting of a shared expert and modality-specific experts for sign and text
tokens. A modality router dynamically activates the corresponding experts, and their outputs are aggregated to form the final
token representations. By enabling modality-dependent expert specialization while preserving a shared expert path, MA-MoE can effectively model the modality differences between continuous sign representations and discrete text
tokens. Extensive experiments on multiple
benchmarks demonstrate that OneSign achieves competitive or state-of-the-art performance on several
benchmarks, highlighting its effectiveness as a unified SLU model.
Datasets are available at : https://github.com/gswycf/OneSign.
Read the full abstract ▾
All authors Shiwei Gan, Yafeng Yin, Xiao Liu, Desibieer Tuerdaken, Lei Xie, Sanglu Lu
Computation and Language · Cross-listed from cs.LG
Algorithmic Harms Associated with Generative Model-Augmented Recommendation Systems
Christine Herlihy, Xumei Xi, Shloka Desai et al. (+2 more)
In this work, we consider algorithmic harms that may arise as generative models are incorporated into machine learning platforms. We argue that existing harm taxonomies and threat models require extension to (1) address novel causal drivers of well-studied representational and quality-of-service harms; and (2) anticipate and mitigate endogenous harms, such as sanitization, which may arise when system inputs are misaligned with the system designer's objectives, or the generative model's inductive priors. To this end, we introduce an expanded taxonomy of algorithmic harms associated with the use of generative models in non-conversational recommendation systems. In addition, we offer a causal analysis of how problematic subsets of the (input, output) joint distribution can arise, in an effort to inform harms detection and mitigation efforts.
Read the full abstract ▾
All authors Christine Herlihy, Xumei Xi, Shloka Desai, Kevin Bannerman Hutchful, Pedro Silva