arXiv:2606. 31543v1 Announce Type: new Abstract: Large language models can produce fluent, internally coherent reasoning traces for abstract reasoning tasks while still being confidently wrong - making selection among candidates, not just generation, the central challenge.
arXiv:2607. 27783v1 Announce Type: cross Abstract: Large Language Models (LLMs) explore problems through chain-of-thought, but this exploration is buried in unstructured prose.
arXiv:2610.01892v1 Announce Type: cross Abstract: Multimodal agents commonly generate free-form reasoning before each action. For small models, limited model capacity can result in lengthy reasoning...
ReSolve is a training‑free inference method that reuses candidate reasoning by selectively moderating generative outputs. It examines existing derivations when candidates disagree or lack a parseable answer, then incorporates new solutions into a bounded loop. On 130 competition‑mathematics problems, ReSolve achieves 100 and 99 correct answers with significantly fewer tokens than eight‑sample self‑consistency, while a controlled ablation shows that visible derivations improve accuracy.
arXiv:2608. 12585v1 Announce Type: new Abstract: Improving reasoning LLMs requires the ability to judge the quality of long reasoning traces for effective reasoning data curation, strong training signals during reinforcement learning, and an in-depth understanding of reasoning behaviors during model performance evaluation.
The paper introduces Tasks over Application Manuals (TAM), a benchmark designed to test long‑horizon procedural reasoning in large language models. TAM uses real‑world tasks from ICD‑10‑CM clinical coding and U.S. federal sentencing, requiring models to follow extensive, rule‑based manuals and perform interdependent steps to produce exact answers. Experiments with GPT‑5 and various prompting strategies show very low exact‑match accuracy—1% for coding and 15.5% for sentencing—highlighting a gap between current benchmarks and the ability to reliably follow complex procedures.
arXiv:2608. 07885v1 Announce Type: new Abstract: Reasoning modes of language models outperform their non-reasoning counterparts on multi-step agentic tasks, but pay a 3-6x premium in output tokens on every episode -- much of it spent re-deriving procedures that are shared across episodes of the same domain.
arXiv:2608. 05643v1 Announce Type: new Abstract: Test-time scaling improves LLM reasoning by using additional inference compute, but wider sampling alone can suffer from diminishing returns: new rollouts often repeat existing answer patterns instead of adding useful reasoning diversity.
arXiv:2509. 14704v3 Announce Type: replace Abstract: Benchmark saturation and training-data contamination increasingly obscure whether reported gains in large language models (LLMs) reflect genuine advances in reasoning or familiarity with recurring patterns in benchmark problems.
arXiv:2604. 11996v2 Announce Type: replace-cross Abstract: Should we trust Large Language Models (LLMs) with high accuracy?
Selecting the best response from multiple small-model samples using a stronger scorer is a simple inference-time strategy, but fails when the small model has already committed to incorrect reasoning paths. PRM guided search avoids this by scoring candidate continuations during generation, but requires a reward model trained with step-level labels.
LOC I (Locator‑Critic) is a training‑free framework that separates visual search from evidence verification in Vision‑Language Models. It uses a Locator agent to propose candidate visual evidence and a Critic agent to assess its relevance, engaging in an iterative refinement loop that progressively improves the evidence until it is sufficient to answer a question. The approach yields state‑of‑the‑art results on several complex visual benchmarks, boosting accuracy for both open‑weight models like Qwen3‑VL and proprietary models such as Gemini 2.5 Pro.
PotARCin expands the ARC benchmark by evaluating abstract reasoning across five dimensions—Definition, Classification, Constrained Generation, Editing, and Inversion—using programmatic generation of new task instances. The study shows a 25‑52 percentage‑point performance gap between standard ARC evaluation and PotARCin, and reveals that multi‑dimensional assessment can reorder models that appear equivalent under single‑metric accuracy. Additionally, a new held‑out set, P‑ARC, demonstrates low model accuracy (1‑8%) across all dimensions, highlighting the need for more comprehensive tests of abstract reasoning.