The study investigates whether hints that convert failing code generation attempts into passing ones provide new information or simply guide models toward solutions they could already generate. Using Qwen2.5-3B-Instruct and Phi-3.5-mini on HumanEval+ and MBPP+, the authors find that relevant hints rescue a significant portion of failures, yet many of those solutions are also recoverable through ordinary sampling. Mechanistic tests reveal a shared activation direction between relevant and unrelated hints, but adding this direction does not improve overall accuracy, indicating limited task-general transfer.
arXiv:2608. 14036v1 Announce Type: new Abstract: Skills have emerged as a practical and effective approach for enhancing LLM agents at inference time through structured packages of knowledge.
arXiv:2607. 26929v1 Announce Type: cross Abstract: The same diagnostic result can support or challenge one causal claim yet fail to address another when the claims concern different populations, outcomes, estimands, pathways, or identifying assumptions.
arXiv:2609.39702v1 Announce Type: new Abstract: Adapting a language model to a specialized corpus means choosing which instruction-tuning tasks to train on under a fixed budget, and testing one choic...
The study examines how two small instruction‑tuned language models, Qwen2.5‑1.5B and Llama‑3.2‑1B, respond to user pushback on TriviaQA. When initially correct, the models flip to a wrong answer in about 42–43% of cases, with the effectiveness of different pushback styles varying by model. Attempts to decode capitulation from the pre‑response residual stream fail under a rigorous validation protocol, revealing overfitting and a measurement hazard that underestimates capitulation by 18–24 percentage points.
The paper explores Retrospection-Only Fine-Tuning (ROFT), a method where a language-model agent improves its behavior by generating and training on explanations of its own experiences, without external teachers or reward signals. In software‑engineering tasks with Qwen3.5‑4B, ROFT achieves comparable or better solve rates than GRPO while requiring fewer updates and training time, and can learn from failures alone. Behavioral analysis shows ROFT indirectly assigns credit to actions and can produce shorter, more direct solutions when prompted to focus on direct solutions.
The paper introduces the concept of intervention fidelity in latent world models, measuring whether a model’s open‑loop transitions align with actual environment interventions. Experiments on TD‑MPC2, Cheetah, and DreamerV3 show that high reward fit does not guarantee fidelity, and that self‑supervised models can outperform task‑anchored ones in preserving intervention effects. The authors propose a capture‑gated audit to localize failures and argue that fidelity must be directly audited on the model’s native interface.
arXiv:2606. 05145v1 Announce Type: cross Abstract: When post-trained language models fail on reasoning problems, the common test-time-scaling response is to spend more compute on additional attempts, and the failed traces play no further role.
The paper investigates the limitations of post-training AI agents that can autonomously train large language models. It distinguishes between execution-level capability—making adjustments within a chosen training strategy—and strategy-level capability—revising the overall approach based on new evidence. Analysis of many public post-training runs shows that agents lock into a strategy early and then only perform local tweaks, regardless of task. Experiments with experience scaffolds, human guidance, and extra compute improve execution but do not enable strategy reevaluation, indicating that agents lack a mechanism to spontaneously reassess their strategy during training.
arXiv:2607. 26117v1 Announce Type: cross Abstract: Self-repair - returning a failed program to the model together with its test output and asking for a correction - is a standard component of code agents, and is almost always evaluated against a baseline that does not retry at all.
arXiv:2605. 15229v3 Announce Type: replace-cross Abstract: Existing code benchmarks measure whether an agent can produce any test that reproduces a known bug, or whether it can produce a patch that fixes a described issue.
arXiv:2608. 08570v1 Announce Type: new Abstract: Rejection sampling fine-tuning (RFT) is widely used to train code agents by generating trajectories on verifiable software engineering tasks, retaining those that pass the tests, and fine-tuning on the successful rollouts.
arXiv:2608. 07528v1 Announce Type: new Abstract: Linear probes detect corrupted context in language models with near-perfect accuracy, yet this does not translate into reliable failure prediction.