Skip to main content
Search

East China Normal University

18 related articles

Subscribe to updates
  1. arXiv

    Vibrational spectra skip per-point DFT: HARMONY anchors on harmonic spectra and returns experimental IR and Raman in 2.12 seconds per molecule

    Predicting experimental vibrational spectra from molecular structure has been caught between theory and measurement: quantum chemistry is accurate but slow, while end-to-end neural networks are fast but bypass explicit vibrational information. HARMONY assembles harmonic spectra from atomic forces and spectroscopic tensors predicted directly by equivariant neural networks, then uses HarmoNet to map them to experimental IR and Raman spectra conditioned on molecular topology and measurement metadata. On the HarmoSpec test set, the full workflow reaches mean cosine similarities of 0.912 for infrared and 0.907 for Raman, with a mean of 2.12 seconds per molecule, roughly 600-fold faster than the evaluated DFT workflow.

  2. arXiv

    GRPO safety alignment loses its signal on hard prompts: Beacon restores a safe direction and cuts average safety error by up to 32.83 points across four models

    Safety alignment of large reasoning models commonly relies on GRPO, but responses sampled for one prompt often share the same safety outcome, so the within-group relative advantage vanishes and the model gets no direction toward safety on persistently unsafe prompts. Beacon pairs safety signal recovery, which filters high-contrast trajectories and directly supervises a safe reference response when the signal collapses, with safety-utility decoupled alignment, which assigns category-aware rewards and applies KL anchoring only to benign prompts. Across four Qwen3 and DeepSeek-R1-Distill variants and eight benchmarks, average safety error falls by 17.69 to 32.83 percentage points relative to the initial policies, with XSTest over-refusal between 0.40% and 1.20%.

  3. arXiv

    Frozen models improve without weight updates: TextualRL turns repeated same-task attempts into context patches, gaining up to 24.4 points on a six-benchmark average

    TextualRL critiques a group of rollouts from the same task under the same context as a unit, rather than an isolated trajectory, and writes reviewed textual patches back into the shared context while model parameters stay frozen. Across five target models and six benchmarks it reaches the highest six-benchmark average for every target, gaining 11.4 to 24.4 percentage points over the no-skill baseline and 0.9 to 3.2 points over the strongest baseline, SkillOpt.

  4. aclanthology.org

    Vision-language hallucinations get a fix at the encoder: three LoRA experts, no retraining per model

    Hallucinations in large vision-language models partly trace to CLIP-style encoders that favor main objects and skip image detail. PLANE attaches three LoRA visual experts to CLIP, each trained on object, attribute or relation captions, and an instruction-aware gate fuses the four image representations at token level. Trained once on LLaVA-1.5 7B, it plugs into other models sharing the same encoder; on LLaVA-1.5 7B the MME hallucination subset total rises from 555.38 to 582.34 and the GIFT hallucination rate falls from 23.39% to 20.30%.

  5. arXiv

    One deblurring network for every camera: Fourier convolution learning lifts GoPro PSNR to 34.02 dB

    A unified motion deblurring network learns blur as a convolution by predicting input-adaptive filters in Fourier space, raising GoPro PSNR from the strong EVSSM baseline's 33.26 dB to 34.02 dB for 8G more FLOPs.

  6. arXiv

    Diffusion LLMs lose accuracy fast at high parallelism; RePara trains two-token blocks and lifts GSM8K from 28.20 to 68.54

    Diffusion language models lose much of their accuracy when committing many tokens per step, and RePara reinforces short consecutive token blocks with group-relative policy optimization, raising LLaDA's GSM8K accuracy at Diff#16 from JustGRPO's 28.20 to 68.54.

  7. A generative ICU twin trained without task labels forecasts 13 endpoints at AUROC 0.83–0.98 and ranks first among 20 predictors

    A 123.5-million-parameter generative model, Didymos, was trained once without task labels on 409,297 intensive care stays from seven databases in five countries and answers prediction questions by sampling a patient's futures. On held-out stays it read 13 primary endpoints from those futures at AUROC 0.83–0.98 and ranked first on average among 20 predictors, including label-trained LightGBM and TabPFN-3.5. Its simulated treatment responses agreed with eight of nine observational contrasts, yet at two hospitals in Russia and China that contributed no training data it needed minutes of continued training on unlabelled local stays to match locally supervised models.

  8. arXiv

    BayesJudge models conflicting human and LLM judgments as posterior distributions, raising expert prediction accuracy to 65.3% on SummEval

    The authors propose BayesJudge, an online Bayesian meta-evaluation layer for conflicting human-LLM judgment streams that outputs a posterior verdict distribution per comparison while estimating rater-specific human confusion matrices and LLM presentation-order bias, using tie-open labels and paired AB/BA order swaps as protocol-level identification instruments; synthetic experiments recover prespecified evaluator parameters, and on SummEval it detects LLM position effects, separates expert from crowdworker behavior signatures without rater metadata, and yields posterior uncertainty that correlates with human disagreement.

  9. arXiv

    WebLoop trains an execution-free Critic with execution feedback, reaching 41.5 Overall on WebRise and 38.9% on WebGen-Bench with a 9B model

    WebLoop jointly optimizes generation, critique, and refinement as roles of one shared policy for functional Web generation: Generator and Refiner outputs receive direct browser-execution rewards, while the Critic never sees execution results and is trained from requirement-level execution labels (discriminability) plus the execution gain induced by subsequent refinements (helpfulness), using a staged reward that first establishes diagnosis and then adds consequence-aware credit. With Qwen3.5-9B, WebLoop reaches 41.5 Overall on WebRise and 38.9% accuracy on WebGen-Bench, improving the base model by 11.3 and 15.4 points.

  10. arXiv

    HarnessSecurity-Bench tests six coding agents: enabling auto-approve raises attack success from 29.2% to 95.6%

    The work builds a ten-mechanism taxonomy across 40 coding agent harnesses and rates 400 harness–mechanism cells (205 confirmed implemented, 83 confirmed absent, 112 unresolved), then introduces HarnessSecurity-Bench, which runs 2,500 trials under GLM-5.2 on 23 tasks across five attack surfaces with separate deterministic oracles for six harnesses, finding that enabling auto-approve raises attack success from 29.2% to 95.6%, that network isolation and read-only mode cut attack success from 57.1% and 8.3% to 0.8% at utility losses of 24.5 and 34.0 percentage points, and that command allowlisting and denylisting reduce attack effects with a small utility loss and a utility gain respectively.