Searcharxiv⌕ Search

arXiv subjects

Gu Simiu

Publications and source records attributed to Gu Simiu.

4 recordsLinked to original sources

Coding Agent Memory Post-Training II: Unlocking the Memory Potential of Pre-trained File Operations for Long-Horizon Tasks via On-Policy Distillation

Language-model agents increasingly tackle long-horizon tasks whose interaction histories exceed the model's active context. They must therefore decide what to write down and what to read back, and when to do either. Recent work has begun to use on-policy distillation (OPD) to make memory control part of the policy, relying on memory tools and compaction protocols that lie outside either model's pre-training. However, this weakens the distillation on both sides: the teacher cannot give useful guidance on an interface it has never used, and the student samples too few meaningful memory actions for that guidance to take effect. To study this problem, we introduce Coding Agent Memory Gym II (CAMG-II), a suite of long-horizon agentic-RL environments spanning Shop, Coding, AutoResearch, General, and Visual. Alongside each environment's native task interface, CAMG-II provides ordinary Bash access and an episode-persistent workspace, enabling the policy to create, revise, search, and reuse files as memory across context replacements. We also introduce Coding Agent Memory Post-Training II (CAMP-II), which trains a single policy jointly across all five environments with fully asynchronous OPD to learn this file-based memory behavior directly from a frozen teacher with a prior for file operations. On CAMG-II, CAMP-II obtains 2.97 times the gain of the strongest OPD baseline. The memory behavior CAMP-II learns generalizes to two external benchmarks, Terminal-Bench 2.1 and SWE-Bench Pro V2, with improved success rates.

cs.LG↗

A Better Spur Should Start From Each Objective

Real-world Multi-Objective Reinforcement Learning (MORL) often suffers from sparse rewards, reward conflicts, and late-stage reward tug-of-war, causing traditional linear scalarization to experience severe metric oscillations. To address optimization conflicts among multiple objectives in real-world deployment scenarios, we propose Multi-Marginal Preference Optimization (MMPO), a fine-grained framework that intervenes at the data, gradient, and constraint levels rather than relying on coarse-grained global scalarization. Specifically, MMPO performs exposure debiasing to mitigate sparse and biased rewards, applies priority-aware orthogonal projection to decouple conflicting gradients, and introduces self-prompted gradient constraints to prevent dominant objectives from overwhelming weaker ones. Experiments on real-world e-commerce datasets show that MMPO improves training stability and consistently achieves better performance across conflicting metrics. Moreover, it generalizes robustly to broader tasks such as ToolRL and code generation, demonstrating its effectiveness as a practical paradigm for multi-objective alignment.

cs.AI↗

Match One, Learn with Graph: One-to-Graph Query Collaboration with Backward Sharing for Object Detection

One-to-one (O2O) matching enables Detection Transformers (DETRs) to perform end-to-end set prediction by assigning each object to a single positive query. However, the strongest classification, center, scale, and overlap evidence for an object is often distributed across multiple queries. This mismatch leaves only the matched owner positively supervised for the object, while other evidence-bearing queries receive no box target for it. We term this query knowledge fragmentation. To exploit such complementary evidence without one-to-many supervision, we propose BS-O2G, a plug-in that builds a sparse prediction-aware graph from decoded features, boxes, and class distributions to organize query collaboration in feature and optimization spaces while preserving the original O2O matcher, positive labels, and objective. One-to-Graph (O2G) calibration propagates relative messages over this graph to consolidate query evidence in the forward pass, whereas Backward Sharing (BS) reuses its transposed detached adjacency to route gradients across persistent query basis vectors without changing the decoder input in the forward pass. Experiments across diverse DETR methods, backbones, COCO, and CrowdHuman show consistent gains and faster convergence with negligible parameter/FLOP growth and modest runtime overhead, supporting graph-based query collaboration as an alternative to expanding positive assignments.

cs.CV↗

Recompute or Reuse? Diagnosing and Mitigating Textual Shortcuts in VLM Self-Reflection

Vision-language models (VLMs) are expected to revise their reasoning when visual evidence changes. Failures to do so are often attributed to insufficient visual attention or contextual inertia, leaving unclear what models reuse instead of recomputing from the current image. We show that evidence-bearing reasoning in a prior chain of thought (CoT) can form a textual shortcut that competes behaviorally with visual recomputation. Across 16 VLMs, a matched counterfactual analysis identifies evidence-bearing content as the most robust carrier of prior-CoT influence. Removing this evidence-bearing content shifts answer preference more than removing length-matched non-evidence context or the final-answer span, with prior control weakening progressively as more stale evidence is removed. Reordering this evidence also weakens prior control, showing that its organization modulates shortcut strength. Beyond the immediate answer, the shortcut can retain residual influence after answer correction: weakening current-image support shifts preference back toward the prior answer, while repeated prior answers and reused premises arise mainly when the shortcut remains active. To limit this influence, we introduce Fresh-State Attention Firewall (FSAF), a training-free intervention that isolates fresh computation from the prior CoT. Across five VLMs, FSAF raises visual update rate from 35.28% to 53.61% and reduces prior-answer rate from 39.22% to 3.67%. Reliable VLM self-reflection therefore requires more than looking again: fresh visual recomputation must be protected from stale textual reuse.

cs.CV↗