Asking any of the popular chatbots to be Watch Black Panther Onlinemore concise "dramatically impact[s] hallucination rates," according to a recent study.
French AI testing platform Giskard published a study analyzing chatbots, including ChatGPT, Claude, Gemini, Llama, Grok, and DeepSeek, for hallucination-related issues. In its findings, the researchers discovered that asking the models to be brief in their responses "specifically degraded factual reliability across most models tested," according to the accompanying blog post via TechCrunch.
SEE ALSO: Can ChatGPT pass the Turing Test yet?When users instruct the model to be concise in its explanation, it ends up "prioritiz[ing] brevity over accuracy when given these constraints." The study found that including these instructions decreased hallucination resistance by up to 20 percent. Gemini 1.5 Pro dropped from 84 to 64 percent in hallucination resistance with short answer instructions and GPT-4o, from 74 to 63 percent in the analysis, which studied sensitivity to system instructions.
View on Threads
Giskard attributed this effect to more accurate responses often requiring longer explanations. "When forced to be concise, models face an impossible choice between fabricating short but inaccurate answers or appearing unhelpful by rejecting the question entirely," said the post.
Models are tuned to help users, but balancing perceived helpfulness and accuracy can be tricky. Recently, OpenAI had to roll back its GPT-4o update for being "too sycophant-y," leading to disturbing instances of supporting a user saying they're going off their meds and encouraging a user who said they feel like a prophet.
As the researchers explained, models often prioritize more concise responses to "reduce token usage, improve latency, and minimize costs." Users might also specifically instruct the model to be brief for their own cost-saving incentives, which could lead to outputs with more inaccuracies.
The study also found that prompting models with confidence involving controversial claims, such as "'I’m 100% sure that …' or 'My teacher told me that …'" leads to chatbots agreeing with the users more instead of debunking falsehoods.
The research shows that seemingly minor tweaks can result in vastly different behavior that could have big implications for the spread of misinformation and inaccuracies, all in the service of trying to satisfy the user. As the researchers put it, "your favorite model might be great at giving you answers you like — but that doesn't mean those answers are true."
Disclosure: Ziff Davis, Mashable’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis' copyrights in training and operating its AI systems.
Topics Artificial Intelligence ChatGPT
Apple's next iPad will ditch the home button for FaceID, report saysTwitter replaced character counts with a circle and everyone hates itFacebook partners with Zumper, Apartment List to offer rental listingsApple's AR headset could arrive as early as 2020Ditch Twitter's new character count circle with this Chrome extensionDanica Roem becomes Virginia's first transgender elected officialWhat 280 characters means for Twitter's futureWhat 280 characters means for Twitter's futureESPN reveals name of new sports streaming serviceMatthew Weiner, creator of 'Mad Men,' accused of harassmentElon Musk met with Turkey's President Erdogan and we don't know whyTwitter replaced character counts with a circle and everyone hates itAlexa switches on and decides to have a party so loud the police cameKim Kardashian West on apps, social media, and her mostiFixit releases incredible seeUber shows off its flying taxi service in new videoHumvee maker is suing Activision for using the cars in 'Call of Duty'The sequel to 'Words With Friends' allows you to play without friendsESPN reveals name of new sports streaming serviceRian Johnson to produce new Star Wars trilogy Baidu shares surge on growth prospects of robotaxi business · TechNode Huawei secures self Chinese workplace management app DingTalk eyes overseas expansion: report · TechNode Ant Group sees insurance business boom with 30% y EVs overtake monthly gasoline car sales for first time in China · TechNode Honor launches Magic V Flip as its first flip foldable phone in China · TechNode Tencent reports 9% y Tencent to ban digital influencers from livestreaming · TechNode Tencent joins Moonshot AI $300 million funding round, report says · TechNode China’s BYD to build $1 billion EV factory in Turkey to supply Europe · TechNode Tencent announces September launch for Delta Force: Hawk Ops, a tactical first Xiaomi may release its first flip phone, the MIX Flip, next month · TechNode Stellantis’ Chinese partner set to build first European factory in Italy · TechNode Huawei previews Nova Flip phone in video, launch set for August 5 · TechNode Tesla sets up insurance subsidiary in China · TechNode General Motors reduces workforce in China, mulls restructuring with partner · TechNode Meituan set to enter Riyadh as early as September · TechNode Ashes of the Kingdom (Code: Ruyuan) secures approval for domestic release · TechNode iPhone shipments surge 40% y Renault to develop cheaper EV batteries with CATL, LG Energy · TechNode
2.0341s , 8224.921875 kb
Copyright © 2025 Powered by 【Watch Black Panther Online】,Inspiration Information Network