Papers for

ai product managers

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

Language models struggle to predict shifting research focus in ai fields

RAP: Research Attention Prediction Reveals Target-Conditioned Evidence Acquisition Biases

Abstract: Large language models (LLMs) increasingly act as research agents, yet their ability to track shifts in research attention is difficult to evaluate because reviews and research ideas lack uniquely verifiable outcomes. We introduce Research Attention Prediction (RAP), a rolling benchmark covering 278 AI/ML fields and 1,390 episodes. At each cut-off, an LLM agent searches a temporally restricted arXiv corpus and predicts the next six months' paper shares across eight frozen research directions. Search generally helps, but all four diagnostic models perform worse than an exact-count exponentially weighted moving average (EWMA) baseline in compositional accuracy. We identify two linked bottlenecks. Under cumulative-history access, State carry-forward outperforms direct Forecast for all four diagnostic models; frozen-evidence replay links a shared component of this reversal to Forecast-oriented policies retrieving a smaller share of recent evidence. Even with exact historical activity, future-specific updating remains limited, with only GPT-5.5 plus reopened Search slightly surpassing EWMA. Fine-tuning on realised outcomes improves Qwen3-4B's forecast Spearman correlation by 0.105 on held-out fields at later origins, with gains also on change-rich episodes.

Wed 9 SeptArtificial IntelligenceComputation and Language
The gist
It is hard to tell if large language models can predict where research will focus next because research ideas don’t have clear, verifiable outcomes. The authors created a way to test this prediction by tracking 278 AI areas and checking how well models guess which topics will be popular in six months. Their tests show that simple past-trend methods often predict better than complex model strategies. They also found that models have trouble updating their predictions based on new information about recent research activity. Training a model on actual past outcomes improved predictions somewhat but challenges remain.
Open → 2609.10092v1

Model training performs best with moderate data from each domain

Everything in Moderation: Per-Domain Coverage Optima and Alignment-Resistant Domain Gaps in Multi-Domain Mid-Training

Abstract: Mid-training, the stage between pre-training and alignment, is where a model's per-domain data composition is typically set by data availability rather than principled design. We ask what that decision buys, and whether a later alignment pass can undo it. In a controlled logical-reasoning setting (Qwen3-8B-Base, with a 4B replication; five semantically rule-disjoint KOR-Bench domains) we train 30 allocations spanning the five-domain simplex, 24 sweep configurations plus six withheld from the fit, at five seeds each. Three findings emerge. First, every domain has an interior coverage optimum: the moderate band ($10\%$-$40\%$) is best for all five domains, and a calibrated permutation test for quadratic interiority gives $P\approx0.010$; the fitted mid-training-only curves, with 8B peaks between $9.9\%$ and $35.1\%$, reproduce for curve shape but not peak location. Second, the gaps survive a fixed-budget alignment pass: compensatory SFT raises 116/120 cells (mean $+4.32\%$) yet bridges $0/240$ pairs at a $5\%$ threshold and $30/240$ at a $10\%$ ratio, an equal-budget uniform control behaves almost identically, and a permutation null would bridge $13.8\pm3.3$ and $77.9\pm8.5$ pairs ($P<0.001$). Third, zero coverage collapses mid-training-only accuracy, though a FineWeb-Edu-only control shows the collapse is commingled with generic drift. An exploratory $θ^*$ allocation attains the largest full-pipeline gain ($+4.36\%$ vs. $+0.80\%$/$+0.64\%$\,pp) but is marginal under Welch test.

Tue 8 SeptArtificial Intelligence
The gist
Training AI models on data from multiple areas is tricky because too much or too little from any one area can hurt performance. The authors found that using a moderate amount—between 10% and 40%—of data from each domain leads to the best results. They also found that trying to fix imbalances later in training doesn’t fully solve the problem, and skipping some domains entirely causes the model’s accuracy to drop sharply. This shows that how data is balanced during mid-training really matters.
Open → 2609.09081v1

Chatbot performance scores vary significantly between api and interface use

API Benchmark Scores Do Not Reliably Transfer to Chatbot Interfaces

Abstract: Benchmark scores are a central currency in model releases: they inform purchasing decisions, shape public trust, and influence policy. Yet, a key assumption underlying benchmark scores is that the model performance measured through APIs faithfully reflects the behavior of deployed systems. We challenge this assumption by auditing ChatGPT, Claude, and Gemini across seven systems and nine benchmarks spanning general capability, social bias, and sycophancy. We find systematic API--interface differences in both accuracy and consistency. On average, API evaluations score 3.4 percentage points higher in accuracy and 2.1 percentage points higher in test--retest agreement than corresponding interface evaluations. For ChatGPT, the performance difference between API and interface access exceeds the API-only difference between GPT 5.3 and GPT 5.4. Put differently, switching access surfaces can degrade performance as much as downgrading a full model generation. We further test whether exposed API controls can reproduce interface behavior by varying system prompts, sampling parameters, and reasoning settings. These controls shift behavior in some cases but do not reliably eliminate the gap. Our findings document a context-validity gap: measurements obtained through APIs do not necessarily generalize to corresponding deployed interfaces, complicating the use of API evaluations as proxies for deployed systems.

Tue 8 SeptArtificial IntelligenceSoftware Engineering
The gist
Models like ChatGPT are often judged by scores from APIs, but these scores don’t always match how the models perform when used in real chatbot interfaces. The authors tested several major models and found that the API versions generally look better in accuracy and consistency than the same models accessed through chatbot interfaces. They also tried adjusting settings in the API to mimic the interface behavior, but this didn’t fully close the difference. This means benchmark results from APIs might not reliably show how well a chatbot will actually work for users.
Open → 2609.08861v1