arXiv subjects

Yanxian Chen

Publications and source records attributed to Yanxian Chen.

5 recordsLinked to original sources

arXiv2026

Precedent Labels and Facts Shift LLM Approval Decisions in Opposite Directions

When a language model decides refund or claim requests under written policies, the system can show it similar past cases. Each past case has two parts: its facts and the decision recorded for it. We separate the two. A model sees a policy, a request and three past cases, and we change either the recorded decisions or the facts, keeping the request fixed. We select requests that are close calls: deciding without past cases, DeepSeek V4.1 Flash approved each in some of four tries and denied it in others. When the past cases are near-copies of the request (its facts with a few details changed), switching their recorded decisions from DENY to APPROVE raises approval by about 30 to 55 percentage points in DeepSeek V4.1 Flash, GPT-5.6 Luna and Claude Haiku 5.5 on policies written for the study, and DeepSeek and Luna show the same on nine public policies paraphrased from real ones. The facts push the other way: shown without their recorded decisions, past cases whose facts support approval make these two models approve less often than past cases whose facts support denial. One reading is that the request looks weaker next to them. This push is not just a correction toward the right answer: in an additional DeepSeek analysis, it was of similar estimated size on requests that human annotators judged approvable and on those they judged deniable. In a separate DeepSeek run on the public policies, when the past cases are near-copies of a different request, switching their decisions moves approval by 9 points instead of 43, while facts that support approval still lower it by 22 points against facts that support denial. Systems that retrieve past decisions should test the two parts separately.

cs.CL↗

arXiv2026

Renegotiable Options Fail to Prevent Hold-Up Between Two Profit-Maximizing Frontier LLM Agents

Do contract terms that protect investment between people also protect it between profit-maximizing language-model agents? In the hold-up problem, a seller can adapt a unit to one buyer at a cost before the price is set. Once that cost is sunk, the buyer can offer almost nothing and the seller will still accept, so a seller who foresees this does not adapt. One remedy is an option: the buyer may buy at a fixed price. In theory the option stops working if the buyer can let it lapse and offer its own price instead. Yet in a laboratory experiment by Hoppe and Schmitz, this renegotiable option raised human investment from 41.5% of pairs without a contract to 67.2%. We had Claude Opus 5.5 and GPT-6.1 Sol, each told to maximize its side's profit, play a one-shot game with the same strategic structure under their four contract regimes and a reference-price control. The agents did what game theory predicts: under the renegotiable option no seller adapted, and under the option that rules out this renegotiation every seller did. In a second experiment in which sellers no longer stated their expected revenue before deciding, removing the profit instruction left investment unchanged under no contract and both options. When we changed the numbers so that the buyer would exercise the option even without adaptation, adapting only added cost, and no seller adapted: the agents did not simply adapt whenever an option was offered or whenever its price covered the cost. Under our protocol, a contract term that raised investment between people did not protect it between agents.

cs.GT↗

arXiv2026

Evaluating Context Compaction on the Questions That Come Later

Language-model agents compact their context for the question at hand, but the compacted context must also serve questions that come later. We evaluate compaction on these later questions, using four corpora that annotate the evidence of several questions per text and recorded agent trajectories, and compare every method with random deletion at the same budget. On MultiRC, keeping the sentences that gpt-6-luna (through the OpenAI Decisions API) judges necessary for the current question raises its accuracy on that question by 11.4 points at 25% of the tokens. On the later questions, however, no method that selects sentences for the current question does detectably better than deleting sentences at random. Deleting words instead of sentences does better: LLMLingua-2, which ignores the question, beats random sentence deletion at the same token count by 12.5 points on the later MultiRC questions. It keeps about twice random deletion's share of the evidence's content words, and at equal content-word retention its gain shrinks to 7.3 points (exploratory). On FEVER, label accuracy can overstate what the kept text supports: the gain of sentence selection is no longer detected when answers are scored against the kept evidence, and the 21.0-point gain of LLMLingua-2 is not shown to rest on the kept text. The ranking also depends on the setting: on Evidence Inference sentence selection beats LLMLingua-2 at a nominal 25% budget, and on $τ$-bench trajectories, whose later tool calls need exact identifiers, it beats random deletion while LLMLingua-2 shows no detected gain. Scores on the question a compaction was made for do not reliably show how it serves later ones.

cs.CL↗

arXiv2026

One Confidence Threshold, Unequal Coverage: Language Gaps in Typed Decision Interfaces and How Translating Inputs Reduces Them

Deployed classifiers often act on a prediction only when its confidence clears a threshold and pass the rest to a person. We ask whether one threshold serves all languages equally on typed decision interfaces, APIs that return a probability for each option of a fixed multiple-choice question. In a confirmatory study with a written analysis plan, three vendors' interfaces classified 1,000 MASSIVE items, each in 51 languages, with a threshold of 0.9. Even when a request is answered correctly in both English and another language, the English version clears the threshold more often, by 8.6-11.7 percentage points on average. Observed coverage, the share of requests handled automatically, is lower than English's in every other language, while the requests the gate handles are rarely wrong. The gap remains when every language is reweighted to the same accuracy, and a standard calibration metric (ECE) does not detect it on one interface. In exploratory analyses, English's coverage stays higher when each language gets its own threshold fitted to the same 5% error target. Translating inputs into English nearly closes the gap at 0.9 but leaves English ahead at that error target. Requests written natively in African languages show a gap in the same direction when English is their human translation. We recommend reporting per-language coverage at the operating threshold.

cs.CL↗

arXiv2020

Realizing Pixel-Level Semantic Learning in Complex Driving Scenes based on Only One Annotated Pixel per Class

Semantic segmentation tasks based on weakly supervised condition have been put forward to achieve a lightweight labeling process. For simple images that only include a few categories, researches based on image-level annotations have achieved acceptable performance. However, when facing complex scenes, since image contains a large amount of classes, it becomes difficult to learn visual appearance based on image tags. In this case, image-level annotations are not effective in providing information. Therefore, we set up a new task in which only one annotated pixel is provided for each category. Based on the more lightweight and informative condition, a three step process is built for pseudo labels generation, which progressively implement optimal feature representation for each category, image inference and context-location based refinement. In particular, since high-level semantics and low-level imaging feature have different discriminative ability for each class under driving scenes, we divide each category into "object" or "scene" and then provide different operations for the two types separately. Further, an alternate iterative structure is established to gradually improve segmentation performance, which combines CNN-based inter-image common semantic learning and imaging prior based intra-image modification process. Experiments on Cityscapes dataset demonstrate that the proposed method provides a feasible way to solve weakly supervised semantic segmentation task under complex driving scenes.

cs.CV↗