When to Ponder: Adaptive Compute Allocation for Code Generation via Test-Time Training

• URL: http://arxiv.org/abs/2601.00894v1
• Date: Wed, 31 Dec 2025 14:49:54 GMT
• Title: When to Ponder: Adaptive Compute Allocation for Code Generation via Test-Time Training
• Abstract summary: We propose PonderTTT, a gating strategy using the TTT layer's self-supervised reconstruction loss to selectively trigger Test-Time Training updates.<n>Our experiments with GPT-2 models (124M to 1.5B) on code language modeling demonstrate that this signal is inference-compatible, requiring no ground-truth labels.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models apply uniform computation to all inputs, regardless of difficulty. We propose PonderTTT, a gating strategy using the TTT layer's self-supervised reconstruction loss to selectively trigger Test-Time Training (TTT) updates. The gating decision itself is training-free--requiring no learned classifier or auxiliary networks; only a single scalar threshold is initially calibrated on unlabeled data and continuously adapted via EMA to maintain target update rates. Our experiments with GPT-2 models (124M to 1.5B) on code language modeling (The Stack v2, teacher-forced perplexity) demonstrate that this signal is inference-compatible, requiring no ground-truth labels. Our Reconstruction Gating achieves 82-89% Oracle Recovery while being fully training-free, significantly outperforming Random Skip baselines (up to 16% lower loss on OOD languages).

Related papers

• Should I Use This Synthetic Dataset for Training? How to Test with Minimal Real Data [73.19814469893088]
  Digital twins and learned world models are increasingly used to generate synthetic data that augment the scarce real datasets available for training artificial intelligence (AI) models in engineering systems.<n>This paper addresses whether training on the augmented dataset improves the true, population-level performance, while consuming as few real test data points as possible.<n>Experiments on a synthetic-data classification task, a DT-aided wireless packet-scheduling task, and a radio-map prediction task show that aeSFT identifies useful synthetic data using substantially fewer real test samples than mean-based sequential testing.
  arXiv  Detail & Related papers  (2026-08-28T07:05:38Z)
• Grokked Models are Better Unlearners [5.8757712547216485]
  Starting from grokked checkpoints consistently yields more efficient forgetting.<n>Post-grokking models learn more modular representations with reduced gradient alignment between forget and retain subsets.
  arXiv  Detail & Related papers  (2025-12-03T04:35:49Z)
• NeuroTTT: Bridging Pretraining-Downstream Task Misalignment in EEG Foundation Models via Test-Time Training [6.030518150035875]
  This paper introduces a two-stage alignment strategy for EEG foundation models.<n>First, we propose NeuroTTT: a domain-specific self-supervised fine-tuning paradigm.<n>Second, we perform self-supervised test-time training on individual unlabeled test samples.<n>Our approach is the first to unify domain-tuned self-supervision with test-time training in large-scale EEG foundation models.
  arXiv  Detail & Related papers  (2025-09-30T14:14:46Z)
• ESLM: Risk-Averse Selective Language Modeling for Efficient Pretraining [53.893792844055106]
  Large language model pretraining is compute-intensive, yet many tokens contribute marginally to learning, resulting in inefficiency.<n>We introduce Selective Efficient Language Modeling, a risk-aware algorithm that improves training efficiency and distributional robustness by performing online token-level batch selection.<n> Experiments on GPT-2 pretraining show that ESLM significantly reduces training FLOPs while maintaining or improving both perplexity and downstream performance compared to baselines.
  arXiv  Detail & Related papers  (2025-05-26T12:23:26Z)
• Test-Time Adaptation with Binary Feedback [50.20923012663613]
  BiTTA is a novel dual-path optimization framework that balances binary feedback-guided adaptation on uncertain samples with agreement-based self-adaptation on confident predictions.<n> Experiments show BiTTA achieves 13.3%p accuracy improvements over state-of-the-art baselines.
  arXiv  Detail & Related papers  (2025-05-24T05:24:10Z)
• Training Language Models to Self-Correct via Reinforcement Learning [98.35197671595343]
  Self-correction has been found to be largely ineffective in modern large language models (LLMs) We develop a multi-turn online reinforcement learning approach, SCoRe, that significantly improves an LLM's self-correction ability using entirely self-generated data. We find that SCoRe achieves state-of-the-art self-correction performance, improving the base models' self-correction by 15.6% and 9.1% respectively on MATH and HumanEval.
  arXiv  Detail & Related papers  (2024-09-19T17:16:21Z)
• Rethinking Classifier Re-Training in Long-Tailed Recognition: A Simple Logits Retargeting Approach [102.0769560460338]
  We develop a simple logits approach (LORT) without the requirement of prior knowledge of the number of samples per class. Our method achieves state-of-the-art performance on various imbalanced datasets, including CIFAR100-LT, ImageNet-LT, and iNaturalist 2018.
  arXiv  Detail & Related papers  (2024-03-01T03:27:08Z)
• Communication Efficient and Provable Federated Unlearning [43.178460522012934]
  We study federated unlearning, a novel problem to eliminate the impact of specific clients or data points on the global model learned via federated learning (FL) This problem is driven by the right to be forgotten and the privacy challenges in FL. We introduce a new framework for exact federated unlearning that meets two essential criteria: textitcommunication efficiency and textitexact unlearning provability.
  arXiv  Detail & Related papers  (2024-01-19T20:35:02Z)
• Enhancing Plasticity for First Session Adaptation Continual Learning [20.62749699589017]
  We introduce Plasticity-Enhanced Test-Time Adaptation in Class-Incremental Learning (PLASTIC)<n>PLASTIC reinstates plasticity in CIL while preserving model stability.<n>It consistently outperforms both conventional and state-of-the-art PTM-based CIL approaches.
  arXiv  Detail & Related papers  (2023-10-17T13:06:39Z)
• Improving Representational Continuity via Continued Pretraining [76.29171039601948]
  Transfer learning community (LP-FT) outperforms naive training and other continual learning methods. LP-FT also reduces forgetting in a real world satellite remote sensing dataset (FMoW) variant of LP-FT gets state-of-the-art accuracies on an NLP continual learning benchmark.
  arXiv  Detail & Related papers  (2023-02-26T10:39:38Z)
• Test-Time Prompt Tuning for Zero-Shot Generalization in Vision-Language Models [107.05966685291067]
  We propose test-time prompt tuning (TPT) to learn adaptive prompts on the fly with a single test sample. TPT improves the zero-shot top-1 accuracy of CLIP by 3.6% on average. In evaluating cross-dataset generalization with unseen categories, TPT performs on par with the state-of-the-art approaches that use additional training data.
  arXiv  Detail & Related papers  (2022-09-15T17:55:11Z)
• CAFA: Class-Aware Feature Alignment for Test-Time Adaptation [50.26963784271912]
  Test-time adaptation (TTA) aims to address this challenge by adapting a model to unlabeled data at test time. We propose a simple yet effective feature alignment loss, termed as Class-Aware Feature Alignment (CAFA), which simultaneously encourages a model to learn target representations in a class-discriminative manner.
  arXiv  Detail & Related papers  (2022-06-01T03:02:07Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.