Frequency-aware remasking lifts diffusion language model test-time scaling
MATH-500 gains reach 3.73 points for LLaDA-8B-Instruct and 7.33 points for Dream-7B-Instruct.
Stories per day · 30 days
Relatedagent-experience1agent-harness1agent-harnesses1benchmarks1browsecomp1claude1diffusion-language-models1evaluation1gpt1gpu-energy1humaneval1inference1latent-reasoning1math-5001ml-agents1mle-bench1multimodal1overfitting1
17 stories, newest first · one card per story
Stories tagged #test-time-scaling
LARC enables continuous latent reasoning for multimodal image generation
LARC uses curriculum SFT and information-gain RL to bridge discrete token and continuous image spaces, delivering 7% GenEval gains and test-time scaling.
MLE-World predicts ML experiment outputs without running training
Finetuned from Qwen3.5-4B, it matched held-out terminal output on 77.9% of programs versus 41.1% for larger prompted models.
Independent review lifts coding-agent accuracy at modest extra cost
A critic review gains 5.9–9.5 points on Terminal-Bench and 3.9–6.6 on SWE-bench Verified at about baseline cost.
ReGE turns recorded agent experience into guidance for a retry
ReGE-Syn raises accuracy 9.9 points for GPT and 8.8 for Claude while cutting guidance-construction tokens versus adapted AggAgent.
ATEST adapts LLM test-time sampling to live GPU energy
On four A100 datasets, ATEST-Online stayed inside the energy budget and tied or led accuracy as load rose.
Harness-delta method finds much agent-harness gain is overfitting
In 21 significant settings, overfitting and test-time scaling explain 54% and 14% of the gain on average.
From Pareto to Preference: Personalized Test-Time Scaling via Amortized Agentic Policy Discovery
PersonTTS reuses prior search to find test-time controllers matching each user's accuracy, latency, and cost goals.
CLIFT: Conformal Self-Verification for Web Agent Training and Test-Time Scaling
CLIFT turns judge feedback into conformal self-verification for web-agent training and judge-free test-time trajectory selection.
CLIFT: Conformal Self-Verification for Web Agent Training and Test-Time Scaling
CLIFT turns judge feedback into reusable rewards and judge-free test-time selection for web agents.
DiVeR: Decision-Critical Verifier Learning for VLA Test-Time Scaling
DiVeR reweights VLA verifier training toward decision-critical states, improving robot task success with little inference overhead.
Improving Test-Time Scaling with Adaptive Looped Transformers
TaH2 adaptively loops tokens that benefit, steepening the AIME accuracy-compute slope 53 percent versus a non-looped baseline.
Improving Test-Time Scaling with Adaptive Looped Transformers
TaH2 adaptively loops layers on hard tokens, steepening AIME test-time scaling by 53% over a non-looped baseline.
CliffCompaction: Cost-Efficient Compaction for Long-Horizon Coding Agents
CliffCompaction cuts long-horizon coding-agent context cost by up to 50% without rewriting retained text.
Sample Count Is Not Enough: Candidate-Generation Strategy Shapes the Energy and Performance of LLM Test-Time Scaling
Study shows LLM test-time scaling cost depends on generation schedule: eight serial A100 calls use ~4.7x more GPU energy than one batched call.
Dynin-Robotics: Omnimodal Unified Diffusion Vision-Language-Action Model
Dynin-Robotics omnimodal diffusion vision-language-action model unifies action, dynamics, and goal prediction, reaching 78.4% success on Franka tasks.
Dynin-Robotics: Omnimodal Unified Diffusion Vision-Language-Action Model
Dynin-Robotics unifies action, goal, and dynamics prediction in one omnimodal masked-diffusion VLA model, reaching 78.4% success on Franka Research 3 manipulation tasks.